This study examined the diagnostic utility of PiB-PET in discriminating between AD and FTLD in a large sample of clinically well-characterized patients, and compared it to the diagnostic performance of FDG-PET, which has an established role in differentiating the 2 diseases.6
We found that amyloid imaging was sensitive and specific in differentiating AD from FTLD, thus fulfilling an important criterion for an AD biomarker.27
FDG-PET is already recognized by US health authorities as useful in this clinical scenario, yet our study suggests that PiB performs at least as well, and has the additional advantages of higher sensitivity and better accuracy and precision of qualitative reads. Furthermore, PiB slightly outperformed FDG in patients with known histopathology. These findings support a role for amyloid imaging in the differential diagnosis of AD and FTLD.
Diagnosing the cause of dementia during life currently relies on correlations between clinical syndromes, topographic patterns of neurodegeneration, and underlying histopathology. The limitations of this approach are increasingly evident, as clinicopathologic studies demonstrate clinical and anatomic overlap between diseases.4,5,7,9
While PiB directly measures molecular pathology, neuroimaging techniques such as MRI and FDG-PET measure the secondary effects of disease on brain structure and function, and may ultimately fail to predict the underlying histopathology when neurodegeneration does not conform to characteristic topographic patterns. For example, 20%–27% of patients with clinically diagnosed AD in our study were judged to have an FTLD-like metabolic pattern, consistent with previous reports that frontal involvement is common in early-age-at-onset AD.28
The majority of these patients were PiB-positive (D), including 1 patient with pathologically confirmed AD (, patient 3).
Visual ratings of PiB scans had a higher sensitivity for AD than visual ratings of FDG, with similar specificity. Based on our a priori quantitative thresholds, PiB had higher sensitivity and negative predictive value and lower negative likelihood ratio, while FDG showed higher specificity, positive predictive value, and positive likelihood ratio (). On ROC analysis, PiB and FDG were found to have similar discriminatory power (nearly identical AUC) but different diagnostic strengths, with PiB showing higher sensitivity and FDG higher specificity at thresholds that optimized overall classification accuracy. These findings suggest a complementary diagnostic role for PiB and FDG. When evaluating a patient with early-onset dementia, the clinician's first imperative is to “rule out” AD, since symptomatic treatments are currently available and novel therapies for AD (many of which target Aβ) are in advanced clinical trials. This could be achieved with PiB with high sensitivity. If the clinical assessment and PiB are at odds, FDG could add value as the more specific diagnostic test, particularly if analyzed quantitatively.
A practical limitation of FDG-PET is that hypometabolism patterns can be ambiguous and difficult to interpret qualitatively.7
Our experienced visual raters achieved good agreement on FDG, but near perfect agreement interpreting PiB studies. Similar results have been reported by another group29
and suggest that, at least in a dementia population, qualitative interpretations of PiB scans are more reproducible than FDG reads. Consistent with previous reports,6
we found that classifying FDG scans quantitatively in reference to a control population enhanced diagnostic accuracy. Several methods for quantifying FDG data to aid with single subject diagnosis are currently available30
or under development,31
and our data suggest that adopting quantification into clinical practice would improve the diagnostic utility of FDG-PET. In our study, agreement between qualitative and quantitative classifications was very high for PiB and more modest for FDG. Amyloid PET may thus be better suited than FDG for the current clinical standard of qualitative assessment, since visual reads are both more accurate and more precise when compared to quantitative methods.
Our study has limitations. The gold standard against which PiB and FDG were judged was clinical diagnosis, and histopathologic confirmation was available only for a subset of patients. However, the clinical assessment was comprehensive and performed by clinicians who are highly experienced in evaluating AD and FTLD, and clinical diagnosis was confirmed in all 12 patients with known histopathology. The rates of PiB-negative AD (11%) and PiB-positive FTLD (16%) in our study are similar to rates of clinically misclassified patients in autopsy series,9,10,32
suggesting that PiB may have outperformed the clinical diagnostic standard in some cases. Additional causes of false-negative PiB scans are likely to include low Aβ burden,33
high amyloid load in the cerebellar reference region,34
and failure of PiB to bind amyloid.35
False-positive PiB scans are likely to represent comorbid Aβ plaques in patients with FTLD or another primary pathology.36
Indeed, one patient in our study with primary FTLD-TDP and comorbid diffuse plaques had a borderline positive PiB scan (by quantitative criteria), though 3 other autopsy-confirmed FTLD patients with early Aβ deposition were visually and quantitatively PiB-negative.
Patients in our study were recruited at an academic dementia center, were required to meet clinical criteria for AD or FTLD, and were relatively young. Our findings may thus not be generalizable to the more clinically ambiguous patients seen in general practice, or to an older population in which the baseline prevalence of amyloid is higher. Though clinically mild, patients in our study all met criteria for dementia, and future studies are needed to compare the performance of amyloid and FDG-PET in the predementia state, when disease-modifying therapies may have the greatest impact. A disadvantage of amyloid PET not addressed in our study design is that it cannot distinguish between different amyloid-positive (e.g., AD vs dementia with Lewy bodies) or amyloid-negative diseases (e.g., FTLD subtypes, psychiatric mimics of FTLD), while MRI and FDG-PET may help in differentiating these conditions.
Though widespread use of PiB is not feasible due to the short half-life of the carbon-11 isotope (20 minutes), new amyloid tracers labeled with fluorine-18 (t1/2
= 110 minutes) have thus far performed comparably to PiB37–39
and could be produced and distributed for clinical use. Future studies are needed to compare the diagnostic performance of amyloid imaging to CSF biomarkers, which have also shown promise in differentiating AD and FTLD.40
While molecular biomarkers will never replace a thoughtful clinical evaluation, their development heralds a new era in which core pathologic features of neurodegeneration can be directly measured and incorporated into clinical decision-making. This will doubtlessly increase diagnostic accuracy during life, a critical first step toward developing effective disease-specific therapies for these devastating illnesses.