Head-to-head comparison of plasma and PET imaging ATN markers in subjects with cognitive complaints

With the increase of the aging population worldwide, cognitive impairment is posing a tremendous burden on our society. In addition to dementia [1], subjective cognitive decline (SCD) [2] and mild cognitive impairment (MCI) [3] are two stages of cognitive decline that frequently occur in advanced age. Alzheimer’s disease (AD), the most common form of dementia, is a multifaceted disease with different pathological and mechanistic substrates. The biological definition of AD, that is, the ATN (Amyloid/Tau/Neurodegeneration) framework, aiming for more precise and early disease identification, has gained substantial attraction in research settings [4]. The ATN biomarkers come in three major forms: cerebrospinal fluid (CSF), plasma, and imaging biomarkers. The development of highly specific immunoassays for CSF and plasma biomarkers and recent advances in the field of positron emission tomography (PET) imaging have largely improved the diagnostic accuracy. Notably, there is accumulating evidence supporting complimentary roles for different sets of biomarkers. For example, a rise of CSF or plasma tau species appears to precede abnormal tau PET imaging during the course of AD [5]. This calls for a better understanding of the reciprocal interrelationships between different biomarker matrices within the ATN framework. There is also an unmet need to standardize and validate a strategic roadmap for routine application of ATN biomarkers in memory clinics [6]. Meanwhile, the assessment of relationship between ATN biomarkers and cognitive symptoms (C) is important given that a clinical-biological rather than a purely biological diagnosis of AD is recommended in clinical settings [7].

In recent years, much has been learned on the diagnostic performances of traditional CSF and imaging A (β-amyloid [Aβ] PET, CSF Aβ₄₂, and CSF Aβ₄₂/Aβ₄₀ ratio), T (tau PET and CSF phosphorylated tau [p-tau]), and N (anatomic magnetic resonance imaging [MRI], ¹⁸F-fluorodeoxyglucose [¹⁸F-FDG] PET, and CSF total tau [t-tau]) biomarkers [4, 7]. Although plasma as a potential source of ATN markers has been increasingly explored to reduce the use of invasive lumbar punctures [8], much validation work remains to be done. Regarding the association between fluid and imaging biomarkers, CSF (Aβ₄₂ for A, p-tau for T, neurofilament light chain [NfL] for N) and imaging (¹⁸F-flutemetamol PET for A, ¹⁸F-flortaucipir PET for T, anatomic MRI for N) biomarkers are reported to be not interchangeable and the optimal approach varies by clinical stage [9]. Recently, plasma p-tau biomarkers (p-tau181, p-tau217, p-tau231) have been suggested to be valid indicators of amyloid and tau PET in clinical and community populations [10‐16], although multiple comorbidities may affect the interpretation of these biomarkers [16]. Meanwhile, similar to the non-commutable correlations between CSF and imaging biomarkers, plasma p-tau (p-tau181, p-tau217, p-tau231) and tau PET (¹⁸F-flortaucipir, ¹⁸F-RO948, ¹⁸F-MK6240) biomarkers are thought to reflect different stages of tau pathology progression [17‐19].

In this study, we present a comprehensive head-to-head comparison of plasma (Aβ₄₂/Aβ₄₀ for A, p-tau181 for T, as well as NfL and t-tau for N) and PET imaging (¹⁸F-florbetapir for A, ¹⁸F-Florzolotau for T, and ¹⁸F-FDG for N) ATN biomarkers in a hospital-based cohort of patients with cognitive complaints admitted to a memory clinic. The Aβ status (positive versus negative) and the severity of cognitive impairment served as the main outcome measures for assessing biomarker performances. The four plasma biomarkers included in the current study are relatively well-established and more readily available than other newly developed ones. The FDA-approved amyloid radiotracer ¹⁸F-florbetapir plays a cornerstone role in the diagnosis of AD [20]. The second-generation tau ligand ¹⁸F-Florzolotau (also known as ¹⁸F-APN-1607 or ¹⁸F-PM-PBB3) could overcome the limitations of first-generation tau PET tracers and reduce off-target binding [21]. ¹⁸F-FDG is the most used PET tracer in nuclear medicine and its accessibility is significantly higher than that of any A and T PET imaging. The current study therefore provides a deeper insight into the comparability of plasma and PET imaging ATN biomarkers, which may be helpful for clinical and research applications. The outstanding strengths of the current study were that all participants were consecutively recruited from a real-life memory clinic, and the aforementioned biomarkers were available to all participants.

The present study has three main findings. First, plasma p-tau181 level was found to be significantly associated with PET imaging ATN biomarkers in the entire study cohort, although this association did not persist when Aβ+ and Aβ− subjects were analyzed separately. Second, we identified four biomarkers (global SUVR value for A, MTL SUVR value for T, NEO-T SUVR value for T, and plasma p-tau181 level for T) with similar and good performance in distinguishing between Aβ+ and Aβ− subjects. Third, we found that an increasing tau burden (as reflected by higher MTL and NEO-T SUVR values on ¹⁸F-Florzolotau PET) and a decreasing glucose metabolism (as reflected by lower metaROI SUVR value on ¹⁸F-FDG PET) were significantly associated with the severity of cognitive impairment in Aβ+ subjects. Glucose hypometabolism, along with elevated plasma NfL level, was also related to more severe cognitive impairment in Aβ− subjects. Taken together, while both ¹⁸F-Florzolotau tau PET and plasma p-tau181 are interchangeable markers for ¹⁸F-florbetapir amyloid PET on detecting the presence of amyloid pathology in unselected patients with cognitive complaints, plasma p-tau181 is preferred in screening considering the cost-effectiveness. Further, our study provides scoping information about the potential usefulness of ¹⁸F-Florzolotau PET and ¹⁸F-FDG PET as markers of clinical severity, and none of the plasma biomarkers included in the current study could be used interchangeably in this regard. Notably, since all participants had cognitive complaints and were recruited from a real-life memory clinic, the current findings may only apply to the symptomatic population.

In addition to the established imaging and CSF markers (Aβ PET imaging and CSF Aβ₄₂/Aβ₄₀ ratio) [40], plasma-based biomarkers have frequently been investigated for their ability to identify subjects with amyloid pathology – with most studies focusing on the plasma Aβ₄₂/Aβ₄₀ ratio. However, reliable studies have shown that the Aβ₄₂/Aβ₄₀ ratio in plasma generally underperforms the established CSF and imaging A biomarkers [41] and is prone to significant analytical variation [42]. By relying on the visual interpretation of ¹⁸F-florbetapir PET images to achieve a dichotomous classification of the Aβ status, our current findings further support the view that the plasma Aβ₄₂/Aβ₄₀ ratio has a limited value as an A biomarker [43]. Additionally, we found that neither plasma nor PET imaging A biomarkers were significantly associated with the severity of cognitive impairment. This finding is consistent with previous observations showing that cerebral Aβ accumulation reaches a plateau during the prodromal stage [44] and that the plasma amyloid biomarker profile does not correlate with cognitive function across the clinical spectrum of AD [45].

The agreement between PET, CSF, and plasma T biomarkers varies widely from 66% to 95% [46] and can be influenced by differences in assays and laboratory procedures [47]. The association of plasma p-tau181 (T) with CSF p-tau181 (T) as well as ¹⁸F-flortaucipir PET (T) and ¹⁸F-flutemetamol PET (A) has been previously reported for Aβ+ subjects and for an unselected sample comprising both Aβ+ and Aβ− individuals – but not for Aβ− subjects analyzed separately [11]. These findings are also consistent with an analysis by Thijssen et al. who reported significant associations of ¹⁸F-flortaucipir PET (T) with both plasma p-tau181 (T) and p-tau217 (T) [48]. In our study, associations of plasma p-tau181 (T) with ¹⁸F-Florzolotau PET (T), ¹⁸F-florbetapir PET (A), and ¹⁸F-FDG PET (N) were examined. However, they reached the threshold for statistical significance only in the entire study cohort. On analyzing Aβ+ individuals separately, our results were not consistent with previous studies possibly because we included only subjects with MCI and dementia due to AD but not those in the preclinical stage, as other authors did [11, 48].

Our data also showed that the plasma p-tau181 level (T) was more closely associated with ¹⁸F-Florzolotau PET (T) than ¹⁸F-florbetapir PET (A), although Thijssen et al. found that the plasma p-tau biomarkers were mainly related to PET imaging A biomarkers (¹⁸F-AZD4694, ¹⁸F-florbetapir) than T biomarkers (¹⁸F-MK6240, ¹⁸F-flortaucipir) [48]. This discrepancy might be because that the present study used a different tau PET tracer (¹⁸F-Florzolotau) and did not include subjects in the preclinical stage. It is worth noting that ¹⁸F-Florzolotau has shown favorable affinity to all types of tau aggregates [21] and is able to detect tau deposition in vivo in the brains of patients with different tauopathies (i.e., three- and four-repeat (3R/4R) tau in AD [49‐51], 4R-tau in progressive supranuclear palsy [31, 52], 4R and 3R/4R-tau in frontotemporal lobar degeneration with tauopathy caused by microtubule-associated protein tau mutations [53]) while ¹⁸F-MK6240 and ¹⁸F-flortaucipir have relatively low affinity for non-AD tauopathies [54‐56].

Plasma levels of p-tau biomarkers have been reported to be strongly correlated with both CSF and PET imaging T biomarkers [11, 14, 15, 46, 48, 57]. In line with these findings, the present study found similar patterns for plasma p-tau181 concentrations and PET imaging results with a second-generation tau tracer (¹⁸F-Florzolotau). Interestingly, these markers also appeared to have an excellent discriminatory ability to distinguish between Aβ+ and Aβ− individuals. Research investigating plasma (p-tau181, p-tau217, and p-tau231) and PET imaging (¹⁸F-RO948, ¹⁸F-MK-6240, and ¹⁸F-flortaucipir) T biomarkers has generally detected an increasing tau burden from the preclinical stage to clinically overt dementia. Aside from evidence that plasma p-tau levels tend to increase in a less pronounced fashion in symptomatic patients [11, 15, 45, 58‐61], the correlations of plasma T biomarkers with the results of neuropsychological testing are generally moderate [17, 45, 48]. The pathophysiological cascade of Aβ- and tau-related processes is not constant during disease progression, that is, as opposed to early in the disease, in the advanced stages such as AD dementia when Aβ fibrils and soluble p-tau levels have stabilized, cognitive decline is associated with the accumulation rate of insoluble tau aggregates [62]. Our data add to previous evidence by demonstrating that only SUVR values on ¹⁸F-Florzolotau PET imaging – and not plasma T biomarker – increased in a stepwise fashion with the increasing severity of cognitive impairment. This result suggests that plasma and PET imaging T biomarkers may convey information that is at least in part not overlapping, with plasma p-tau181 concentrations being more closely related to Aβ pathology and tau PET imaging findings being mainly a reflection of the cognitive impairment severity [46]. However, this possibility requires confirmation given the non-linear increase in plasma p-tau181 concentrations during the course of AD [63]. Another prospective research with different T biomarkers from those of our study consistently indicated that the soluble tau as reflected by elevated plasma p-tau217 and the insoluble tau aggregates as reflected by elevated tau PET (¹⁸F-RO948 and ¹⁸F-flortaucipir) signals, are optimal predictors for longitudinal tau accumulation in the brains of patients with AD at preclinical and prodromal phases, respectively [18]. Future work in this area should also validate different plasma T biomarkers, which may have differential roles for identifying amyloid pathology [64]. Moreover, since comorbidities such as chronic kidney disease are reported to have a non-negligible impact on the interpretation of plasma p-tau181 and p-tau217 levels [16], further studies exploring their potential impact on tau PET biomarkers are warranted.

Given that neurodegeneration is the final consequence of various pre-existing pathological alterations, research has generally explored the association of N biomarkers with the severity of cognitive impairment and clinical trajectories over time [65]. However, ¹⁸F-FDG PET as an imaging N biomarker also shows diagnostic value among patients present to memory clinics with an uncertain diagnosis [66]. In line with prior studies [14, 67], metaROI SUVR value on ¹⁸F-FDG PET imaging was the only N biomarker capable of distinguishing between Aβ+ and Aβ− individuals, although it underperformed both A and T biomarkers. There is also evidence that, different from other N biomarkers, glucose hypometabolism on ¹⁸F-FDG PET may predict a steeper cognitive decline trajectory; therefore, the traditional classification of ¹⁸F-FDG PET imaging as an N biomarker has been put into question [68]. Interestingly, we found that ¹⁸F-FDG PET outperformed NfL – a plasma N biomarker – in reflecting the severity of cognitive impairment in Aβ+ individuals. However, the plasma NfL level was superior to ¹⁸F-FDG PET as a marker of disease severity in Aβ− individuals – a finding which calls for additional investigations.

While we are not aware of any other study that has provided a head-to-head comparison of plasma and PET imaging ATN biomarkers in relation to the presence of amyloid pathology and the severity of cognitive impairment across the AD spectrum, several design limitations should be acknowledged. Since this single-center investigation was cross-sectional, it is not possible to establish the causal nature or the directionality of the observed associations. We did not obtain longitudinal measures of cognitive impairment, which restricts the prognostic impact of our findings. Meanwhile, the sample size of the final study cohort was limited, and attention needs to be paid to potential sources of bias. We have only enrolled patients presented to a memory clinic and consequently we were unable to include subjects in the preclinical stage. Our findings may be most applicable and generalizable to those with MCI or dementia due to AD. Another limitation is the uneven distribution of amyloid pathology, resulting in more participants within the Aβ+ group. The uneven distributions of different severities of clinical cognitive impairment, a common drawback of the serial-enrollment design when the study sample size is limited, also requires attention. The present analysis did not include measurements of recently developed plasma T biomarkers (i.e., p-tau217, p-tau231), as well as of neuroinflammatory markers. Because the availability of ¹⁸F-Florzolotau PET imaging is still limited, our results are not conducive to establishing a definitive equivalence between this imaging modality and the combination of plasma p-tau181 and ¹⁸F-FDG PET. Besides, two atlases were used for PET imaging analysis, which may have attenuated some of the results although preceding data rendered such effects likely to be minimal [69]. Last but not least, we took the visual assessment of ¹⁸F-florbetapir PET imaging as a ground truth for Aβ status. As semi-quantitative binary cutoffs (i.e., a global SUVR greater than 1.1 indicates positive Aβ accumulation) [34, 70] have been recommended for ¹⁸F-florbetapir, it is necessary to further replicate our findings using semi-quantitative measurements as the ground truth.

The results from the present study raise the possibility that ¹⁸F-florbetapir PET imaging (A), ¹⁸F-Florzolotau PET imaging (T), and plasma p-tau181 (T) can be considered as interchangeable biomarkers in the assessment of Aβ status in both MCI and dementia due to AD. The findings on ¹⁸F-Florzolotau PET (T) and ¹⁸F-FDG PET (N) could also serve as imaging markers for the severity of cognitive impairment.