Currently, numerous studies have focused on exploring the efficacy of plasma biomarkers in AD prediction, diagnosis and monitoring. However, it is still difficult to draw definite conclusions about the changes in plasma biomarkers due to the inconsistency in different studies. In the present study, we found that plasma biomarkers, including Aβ42, Aβ40, Aβ42/Aβ40, t-tau and NfL showed significant differences between AD patients and healthy controls. Interestingly, our most striking finding was that the plasma NfL level was positively correlated with CDR score, indicating that NfL was probably the most promising candidate biomarker to reflect disease severity and monitor the disease progression, although it may not be a specific marker for AD. To our knowledge, there are few domestic studies detecting these plasma biomarkers together in a large cohort, especially in analyzing their correlations with AD typical biomarkers[16-17]. In the present study, a comparatively large-scale AD patients of blood biomarkers, as well as CSF core biomarkers and amyloid PET were analyzed, and interestingly, we found the plasma Aβ42 and Aβ40 were positive associated with CSF Aβ42/Aβ40, suggesting that plasma Aβ can partly reflect brain Aβ status and be superior to other plasma biomarkers. Besides, we compared the efficacy of plasma biomarkers for the diagnosis of AD by plotting ROC curves. According to logistic regression analysis, after introducing age and APOE genotype into the combined model, including Aβ42, t-tau and NfL, the diagnostic performance reached the maximum value. It is encouraging the model can be applied to perform initial screening in population with a high risk of AD, which might effectively reduce the application of lumber puncture and PET examinations in clinical practice.
In this study, before adjusting for age, sex, and APOE genotype, plasma Aβ42 levels were significantly lower in AD patients, while after controlling for these factors, the difference became insignificant. Previously, the change in plasma Aβ42 levels in AD patients is still controversial [18-19]. Approximately 30%-50% of blood Aβ comes from a brain-to-blood transport mechanism, and there is a dynamic equilibrium between peripheral blood and the CNS[20]. It is possible to measure the level of Aβ42 in plasma to reflect pathological changes in the brain to some extent. Based on this, we assessed the correlation between plasma Aβ42 levels and brain Aβ pathology, including CSF Aβ42 and amyloid PET. Interestingly, we found the plasma Aβ42 and Aβ40 were positive associated with CSF Aβ42/Aβ40, suggesting that plasma Aβ can partly reflect brain Aβ status and be superior to other plasma biomarkers, which was consistent with some previous results[21]. Moreover, plasma Aβ42 showed negative correlation with CSF p-tau, suggesting that, to some extent, plasma Aβ can partially reflect the tau pathological changes of brain, which was consistent with the current research view that there lies in close relationship between Aβ and tau pathology In addition, we found that plasma Aβ42 was shown no correlation with Aβ deposition upon amyloid PET evaluation in our study, which was different with some previous studies[22]. Given that the results reflected different forms of Aβ, the former reflects Aβ peptide and the latter reflects the fiber form, it may reasonable that Aβ in plasma cannot represent Aβ accumulation in brain parenchyma. Meanwhile, the small sample sizes of CSF and amyloid PET in this study may not reflect the real relationship between plasma Aβ42 and brain pathology. Thus, to confirm the real trend of plasma Aβ42 in AD patients and the utility of monitoring the pathological changes in brain tissue, head-to-head comparisons based on large-scale prospective studies are necessary, and unified and standardized inclusion criteria, detection and analysis methods are urgently needed.
For Aβ40, the results are mixed, and the meta-analysis by Olsson et al. suggested that there was no difference between AD patients and controls[23]. In our study, compared with the CN group, Aβ40 was significantly higher in the plasma of AD patients, contrary to the trend of Aβ42. Until now, the diagnostic utility of CSF Aβ40 alone was limited, while it is usually used as a reference peptide that could explain the difference in CSF concentration between individuals and the difference in preanalytical processing of the samples, which otherwise may lead to false-positive or false-negative results using Aβ42 alone[24]. In addition, after considering sex, age and APOE genotypes, plasma Aβ40 was correlated with CSF Aβ42/Aβ40, which is consistent with previous studies that plasma Aβ40 can reflect Aβ pathological in the brain to some extent[22].
Compared to healthy controls, plasma Aβ42/Aβ40 was significantly decreased in AD patients and was in line with most previous studies, which showed a higher degree of consistency than single Aβ peptide even using different detection methods[7, 25]. Decreased Aβ42/Aβ40 levels demonstrated value in reflecting pathological changes of AD and predicting cerebral Aβ status, whether regarding amyloid PET or CSF Aβ as a positive reference[26-28]. Furthermore, the ratio in plasma appears to be associated with an increased risk of progression to AD dementia[29]. This means that if using the plasma Aβ42/Aβ40 ratio as a prescreening tool for the most likely AD population, the necessity of amyloid PET inspection may be reduced, which is of great significance for clinical application and scientific research. However, these findings should be taken cautiously, in our study, the aforementioned results were not obtained, and we speculated that the sample size and measurement method cannot be ruled out as the reason; thus, further validation based on large-scale AD cohort studies is necessary.
Plasma t-tau was significantly increased in AD patients compared with the healthy controls in this study, similar to most published results [17, 30], while several studies also found that plasma t-tau showed opposite changes or no significant changes between AD patients and controls [29, 31]. Nowadays, the results of plasma t-tau in AD patients were still contradictory. Mattson et al. found that higher plasma t-tau was associated with AD dementia and showed a significant correlation with worse cognition, more atrophy and more hypometabolism during follow-up in the ADNI study[30]. However, in our study, plasma t-tau showed no significant correlation with cognitive function and can not reflect the levels of t-tau in CSF, which is also in line with some previous findings [30, 32].Besides, increasing evidence supports that plasma p-tau is a promising biomarker for AD, which can not only be associated with CSF p-tau but also correlates with amyloid PET uptake and discriminates AD dementia from non-AD neurodegenerative disease, partly reflecting AD pathology and disease severity[33-34]. In this study, due to the low concentration in the samples and methodology limitations, we did not measure p-tau in plasma. In the next study, we will increase the sample size to reevaluate whether plasma t-tau could predict cognitive function and AD pathology, and assess the utility of plasma p-tau for discriminating patients with AD from controls.
Emerging evidence supports that NfL is a sensitive and promising biomarker for neurodegenerative disease, which is released into CSF and plasma after axonal damage [35-36]. In the latest research, Yakeel T Quiroz et al. found that plasma NfL increased with age and began to differentiate PSEN1 E280A mutation carriers from noncarriers as early as age 22 based on a large kindred study of patients with AD[37]. In agreement with these results, we found that the plasma NfL concentration was associated with age, and became significantly elevated in the plasma as the individuals became older. Meanwhile, the results of correlation analysis showed that the level of plasma NfL was not associated with CSF core biomarkers, which is not consistent with some previous study that plasma NfL levels correlated significantly with CSF core markers[38-39]. Based on these results, plasma NfL is a sufficient sensitive marker to detect axonal damage in the early stage of AD, while it is not a specific indicator to reflect the typical pathological changes of AD [40]. Interestingly, according to CDR score, subgroup analysis was performed, it was found that plasma NfL levels increased with the severity of the disease, supporting that the NfL level is a marker of progressive myelinated axonal damage that develops in the mild to late stage of AD. Combined with a recent longitudinal study, which demonstrated that the plasma NfL level in participants who developed AD rose at a constantly higher rate than that in CN subjects and increased as early as 10 years before an AD diagnosis[41], plasma NfL levels may be a stable and useful indicator for disease identification and for monitoring disease progression.
Finally, we compared the efficacy of markers in plasma for the diagnosis of AD by plotting ROC curves. The results showed that the plasma NfL level has the best performance compared with other single indicators. Combined with previous studies, due to the poor specificity of NfL for discriminating AD and other neurodegenerative disorders, we tried to combine multiple indicators to verify whether the diagnostic efficacy of plasma markers could be improved. The results showed that when NfL, t-tau and Aβ42 levels were combined, the AUC reached its maximum (AUC = 0.86, sensitivity = 79.41%, specificity = 81.05%). Moreover, even though all indicators (Aβ42, Aβ40, t-tau, NfL, and ratio of Aβ42/Aβ40) were introduced into the model, the diagnostic efficacy was not significantly optimized. Furthermore, after introducing age and APOE genotype into the combined model, the diagnostic efficacy was significantly improved (AUC = 0.88, sensitivity = 82.84%, specificity = 81.69%). This finding is encouraging because the model can be used to perform initial screening in people with a high risk of AD or probable AD, which can effectively reduce lumber puncture and PET examinations.
Notably, the results suggested that these plasma indicators were promising for discriminating AD and CN subjects. The use of a peripheral biomarker panel with low cost and invasiveness as an initial screening funnel to identify people who should undergo further examination, such as PET imaging or CSF testing, might be a critical step forward because subjects at an early stage of AD are hard to identify. In addition, given the low concentration of markers in plasma and the limited detection sensitivity, methodological limitations still exist. To minimize possible experimental errors and promote the detection accuracy of plasma markers, we unified the detection method for each marker via Simoa technology.
Of course, the study still has some limitations, including the following aspects. First, the study was a retrospective study without longitudinal follow-up to determine the trajectories of plasma markers from CN individuals to those in different AD stages. Second, in this study, due to the expensive or invasive characteristics of PET or lumbar puncture, we failed to obtain results for all participants; thus, some individuals were not able to contribute standard pathological results to the group data for further analysis. Next, there is a lack of patients with non-AD dementia to compare the difference in plasma markers and verify the ability of the panel to distinguish AD dementia from other forms of dementia.