We provide a comprehensive overview of the use of fluid-based biomarkers in AD trials starting between 01-01-2017 and 01-01-2024, evaluating the frequency and purpose of utilization of these biomarkers as endpoint. Overall, 44% of the trials used fluid-based biomarkers as an endpoint to monitor either biological treatment effects and/or target engagement. Biomarkers to show biological treatment effects were employed as a primary endpoint most often in phase 2, and the percentage of fluid-based biomarkers as primary endpoint decreased in phase 3 trials. This was expected given that the purpose of phase 3 trials is to show a clinical benefit. CSF and blood-based endpoints were used approximately equally, which was unexpected considering the burden of repeated CSF sampling. The classical pathologic AD ATN markers, Aβ, pTau isoforms and tTau, were used most often in CSF and not yet in blood. Use of target engagement markers was limited (26%). Furthermore, there were several drug target classes that did not include any fluid-based target engagement markers, including drugs targeting neurotransmitter receptors, neurogenesis, vasculature, epigenetic regulators, proteostasis, and the gut-brain axis. We also show there are several trials that have reported biomarker findings, without positive clinical findings. Combined, these findings show that there is an unused potential for the use of fluid-based biomarkers and a need for novel fluid biomarkers to fully capture the complex biology of the disease and for further implementation in clinical trials.

Treatment effect biomarkers are needed in disease-modifying trials

We show that 44% of the trials included in this study used fluid-based biomarker endpoints, mostly in phase 2 (54%). Throughout all phases, fluid-based biomarkers were mostly employed as exploratory endpoints. However, as the goals of the field shift to primary prevention, inclusion of biomarkers as primary endpoints is crucial, as determining efficacy based purely on cognitive endpoints will become more challenging [44, 45]. Specific considerations on biomarker application in clinical trials are shown in Fig. 4. Ideal fluid biomarkers have been investigated in the context of both the disease, the drug mechanism, and are technically mature.

Fig. 4

Framework biomarker development and application in clinical trials

The classical AD biomarkers [46] were employed frequently and can reliably be measured in CSF and blood [6, 17, 47,48,49,50]. Changes in these biomarkers could be indicative of treatment effects and disease modification. Additionally, we show that GFAP was used in only 10 trials. Recent studies show that especially plasma GFAP rather than CSF GFAP correlates with Aβ pathology and has a high prognostic value [22,23,24, 51]. Therefore, GFAP might be additive to Aβ and pTau, which are often directly modified by the drugs, as a marker of disease modification in the early stages of AD. While treatment effect on GFAP might not be required for FDA approval, it can be viewed as strong evidence for a downstream effect of disease modification because it is not related to the target. A decrease in GFAP could be indicative of slowing disease progression, but longer follow-up is needed to prove this hypothesis and better understand the mechanism [23]. The donanemab and lecanemab trials showed its responsiveness in patients with early symptomatic AD [26, 27].

Future of both CSF and blood-based biomarkers

One-third of the trials that employed fluid-based biomarker endpoints used a combination of both CSF and blood-based biomarkers, another third only used CSF biomarkers and the last third used only blood-based biomarkers. CSF and blood-based biomarkers both have their own advantages. The advantage of using CSF is its close proximity to the brain, thereby likely providing a reflection of ongoing brain pathology without peripheral effects on the biomarker levels [52]. Blood-based biomarkers are susceptible to metabolism and excretion interference but offer a low invasive alternative to CSF [53]. This obvious advantage allows for easier serial analyses which can thereby promote trial participation.

Fluid-based target engagement markers are not used to the fullest potential

Fluid-based target engagement can be useful to demonstrate preclinical to clinical translation in early drug development (Fig. 4). Proving target engagement as early as possible saves time and resources, for example by enabling a Bayesian trial design [54], risk of failure of later-stage trials can potentially be reduced [55]. When target engagement is demonstrated, but no effect on cognition is found, this could indicate that the target is not fit or a different approach is needed. This is demonstrated by the results of Semorinemab, gosuranemab, and tilavonemab (Supp. Table 3; Table 3) [36, 38, 39]. The biomarker findings indicate there is target engagement, however, this is not translated to disease modification and clinical benefit. These drugs have not been further investigated in larger trials.

Only 7 of the 36 trials that investigated drugs targeting synaptic plasticity/neuroprotection employed target engagement markers, even though multiple markers are available. NRGN, a post-synaptic marker was used most frequently, i.e. 5 times. CSF presynaptic synaptosomal-associated protein 25, vesicle-associated membrane protein-2, and growth-associated protein 43 (GAP43) have recently been described as synaptic biomarkers, and can likely provide information on presynaptic integrity [56,57,58,59]. β-synuclein, which can be measured in both CSF and blood and relates to Aβ-pathology, could also be a useful marker to investigate synaptic integrity [60,61,62]. Due to limited data from trials, it is unknown which of the synaptic markers, or a panel could detect treatment effects on synaptic function. Therefore, it is too early to conclude whether pre- or post-synaptic markers are more appropriate, or have added value for showing target engagement. By including them in trials at the early stages we also generate insights into the performance of the biomarkers and which assays are suited for clinical trial interpretation. Using both pre- and postsynaptic markers in combination can generate insight into the synaptic health of the neurons and target engagement.

Strikingly, there are several target classes not using any fluid-based marker, even though there are markers available for some of these. These include drugs targeting neurotransmitter receptors, neurogenesis, vasculature, epigenetic regulators, proteostasis, and the gut-brain axis. Potential markers for medication focusing on vasculature include vascular cell adhesion molecule-1 (VCAM-1) or intercellular adhesion molecule-1 (ICAM-1), markers related to vascular endothelium [63, 64]. Research indicated that higher levels of VCAM-1 and ICAM-1 were associated with increased Aβ and tau pathology [65]. Soluble platelet-derived growth factor receptor-β, a pericyte marker, or vascular endothelial-cadherin (VEC), a marker for endothelial injury can give information on blood–brain-barrier integrity [66, 67]. VEC concentrations are increased in preclinical AD [67] and this marker could therefore be implemented in trials focusing on early AD stagesfor potential use as both a target engagement marker or to demonstrate disease modification even in these early stages.

The importance of regulatory status for biomarker implementation in clinical trials

The guidance documents of the US Food and Drug Administration (FDA) and the European Medicines Agency (EMA) focus on cognitive effects to determine efficacy in AD clinical trials. However, both also indicate that biomarkers should be used to support disease modification [68, 69]. Recently, the FDA has shown to be increasingly open to the use of biomarkers as surrogate endpoints in neurodegenerative diseases. The agency gave Accelerated Approval of aducanumab and lecanemab for AD, and tofersen for amyotrophic lateral sclerosis (ALS), indicating that these treatments demonstrated ‘an effect on a surrogate endpoint (Aβ PET for AD and CSF NFL for ALS) that is reasonably likely to predict a clinical benefit to patients’ [70,71,72,73]. While the use of fluid biomarkers for AD trial population enrichment has official support from these agencies, such approvals are not available for biofluid endpoints [74, 75]. In a recent support letter, the EMA recommended the monitoring ability, intra-individual variability, population variability, and behavior over time of the biomarkers needs to be established for such qualification approval [76]. The qualification approval of fluid-based biomarkers as clinical trial endpoints can advance their role in treatment evaluation. Moreover, approval by regulatory agencies could provide an incentive for big pharma to implement them in larger trials also as primary endpoint. Here we show only 27% of trials sponsored by big pharma employed fluid biomarkers. With more big pharma implementing biomarkers, the field can gain insights into their specific uses within a trial setting and more data on the biomarkers over time will be generated, especially if trial data is published and shared. This aid in the interpretation of biomarker results in relation to clinical endpoints to establish insights into the effect sizes required for clinical benefit.

Future perspectives

With current technological advances, we are able to quickly analyze a significant amount of proteins with higher accuracy to establish new potential markers. A recent study on the CSF proteome identified new non-amyloid-related endpoint markers [77]. This could aid with the development of biomarkers for target classes where there is no or few biomarkers available and offer new ways to measure general biological effects. With improved technologies the biomarkers can be combined and multiplexed, which allows for a large number of proteins to be measured quickly making trial analyses easier [53, 78, 79]. Furthermore, advances in technological sensitivity will aid the development of novel blood-based biomarkers.

The aducanumab, donanemab, lecanemab and gantenerumab trials (Table 3) can give insights into the relation of several biomarkers with cognitive outcomes. This can bring us closer to the holy grail of surrogate biomarkers. Based on the study findings, CSF pTau181 and plasma pTau217, pTau181, and GFAP seem to have the most potential for surrogacy biomarkers in the amyloid pathway. Surprisingly, NFL did not respond to treatment in those trials, while in multiple sclerosis (MS) trials this is a very good marker for treatment monitoring, and evidence is developing for ALS [80,81,82]. Potentially, NFL effects are further downstream in AD compared to MS and ALS, thus effects take longer to be visualized in AD. Moreover, there is a bigger relative increase compared to healthy age-matched controls in MS and ALS compared to AD [83]. In order to substantiate the use of fluid-based biomarkers as surrogate endpoints in AD trials, understanding the relation of biomarker dynamics, e.g. if biomarker reduction below a certain threshold, within a critical time-window or between different groups (e.g. sex, APOE4 carriers) is required, will be key in the implementation as surrogate endpoints.

Limitations of the study

Not all biomarker analysis plans may be registered on clinicaltrials.gov. Often there is a significant time period between trial initiation and the end of trial date. Development of biomarkers may have significantly changed nearing the end of trial, and analysis plans may be finalized towards the trial completion, while novel biomarkers can also be included in post-hoc analysis. This might lead to underrepresentation of certain biomarkers in this analysis.

A downside of the biofluid biomarker field is the variable level of validation of the assays, ranging from very standardized high throughput to explorative assays with high CV. Therefore the power of the studies, and risk of false positive and false negative findings is difficult to estimate.

A lot of the studies included in this analysis are not yet finished, so it cannot be investigated if trials with biomarkers have higher success rates. As more successful trials are needed to definitely determine the future role of certain biomarkers, it would be interesting to see the final results of the biomarkers in relation to clinical outcomes. Several trials reported changes in biomarkers and some of these together with clinical effects. Whether biomarker thresholds for clinical benefit can be established, becomes a testable hypothesis as more data is becoming available. To facilitate this, the assay standardization efforts are very important because it enhances the comparability. In addition, a structured re-analyses, engaging the trial investigators, using meta-analyses techniques that account for assay, design, population and trial mechanisms differences could be used to estimate these response relationship and possibly thresholds in a similar manner as has been done for amyloid PET [84].