Abstract
Introduction
Cerebrospinal fluid (CSF) glial fibrillary acidic protein (GFAP) is a neuroinflammatory marker reflecting reactive astrogliosis and is measured regularly in clinical practice. However, its diagnostic utility in differentiating dementia subtypes remains unclear. This study aimed to evaluate differences in CSF GFAP concentrations and its associations with markers of disease severity and amyloid pathology.
Methods
We conducted a retrospective cohort study using three datasets encompassing a broad range of dementia diagnoses. Included variables were CSF GFAP, β-amyloid 42 (Aβ42), the Aβ42/Aβ40 ratio, Mini-Mental State Examination (MMSE) scores, and time from sampling to death.
Results
A total of 1,345 individuals were included. In Parkinson’s disease dementia (PDD) and Lewy body dementia, GFAP levels were similar (p > 0.05). Lower levels were observed in PDD compared to early-onset Alzheimer’s disease (EAD), late-onset AD (LAD), and vascular dementia (VaD) (all p < 0.05); however, the discriminative performance was low-to-moderate: PDD versus LAD (AUROC = 0.74, CI = 0.64–0.84, p < 0.001), VaD (AUROC = 0.71, CI =0.61–0.81, p < 0.001) and EAD (AUROC = 0.59, CI = 0.47–0.71, p = 0.13). Associations were seen with MMSE in mixed AD and VaD (p = 0.027), but not in the other diagnostic categories. GFAP levels did not differ between subjects grouped according to Aβ42/Aβ40 status (p > 0.05).
Conclusion
CSF GFAP did not exhibit clinically relevant diagnostic or prognostic value in dementia. Further studies are needed to clarify its role in PDD.
Keywords: Glial fibrillary acidic protein, Aβ42/40, Alzheimer’s disease, Parkinson’s disease, Vascular dementia
Introduction
The two most common causes of dementia are Alzheimer’s disease (AD) – defined by the intracerebral accumulation of β-amyloid plaques and neurofibrillary tangles, and vascular dementia (VaD) – a heterogeneous disease characterized by vascular lesions in the central nervous system (CNS) [1, 2]. Parkinson’s disease (PD) is characterized by the accumulation of alpha-synuclein, leading to the formation of Lewy bodies, which might affect cognition – resulting in Lewy body dementia (LBD) or Parkinson’s disease dementia (PDD) [3, 4]. Previous studies suggest overlapping pathologies between AD, VaD, and PDD [3, 5, 6]. In common, dementias are characterized by intracerebral protein aggregation, neuroinflammation and neurodegeneration; interrelated processes which are the neurobiological hallmarks of the disease [7]. In the neuroinflammatory process, reactive astrogliosis, a response by astrocytes to harmful chemical agents and molecules inducing CNS injury, has been found to be a key contributor to dementia progression [8, 9]. Glial fibrillary acidic protein (GFAP) has been acknowledged as one of the most trusted markers of reactive astrogliosis [8, 10]. A meta-analysis showed elevated levels of cerebrospinal fluid (CSF) GFAP in AD patients at baseline were related to a worse prognosis in Mini-Mental State Examination (MMSE) [9]. Further, associations between CSF GFAP and the Aβ42/40 ratio have been seen in AD cases, suggesting a concurrent neuroinflammatory response with amyloid plaque deposition [11]. Elevated levels of the marker have also been found in VaD, PD, and FTD [9, 12].
In Parkinson’s disease (PD), CSF GFAP has exhibited ability to accurately predict cognitive deterioration [13, 14]. In traumatic brain injury, high levels of plasma GFAP the days after the event were associated with an elevated 6-month mortality risk [15].
To our knowledge, no previous study has investigated the levels of CSF GFAP, nor the marker’s diagnostic and prognostic performance, across a wide range of dementia diagnoses. Thus, we aimed to assess these gaps of knowledge in the Swedish dementia registry (SveDem); a large national health registry with a broad referral base, containing information from multiple dementia groups. We hypothesized that higher CSF GFAP levels at time of sampling would correspond to a worse prognosis in terms of survival and cognition, and that an association with Aβ42 and Aβ42/40 concentrations would be seen in the AD groups.
Methods
Cohorts
This was a retrospective cohort study, where three sources of information were used. The first was a complete set of archived data on CSF Aβ1-42, CSF Aβ1-40, and GFAP measurements made in clinical practice in the Clinical Neurochemistry Laboratory at the Mölndal site of Sahlgrenska University Hospital, Sweden from January 1, 2010, to October 22, 2022. This laboratory is currently the sole provider of clinical routine CSF GFAP measurements in Sweden.
The second source was SveDem, a national quality registry started in 2007 collecting data on patients under dementia investigation and care reported by primary and specialist care providers in Sweden [16]. All memory clinics and 78% of the primacy care centers in Sweden were connected to SveDem at time of data extraction [17]. Information on clinical diagnosis, MMSE scores, patient age and gender were collected. In the registry, dementia diagnoses classified according to ICD-10 and DSM-V [18, 19] are coded as early-onset AD (EAD, <65 years old), late-onset AD (LAD, ≥65 years old), VaD, AD in combination with VaD (mixed dementia, here MIX), LBD (according to McKeith criteria [20]), PDD (according to Movement Disorder Society Task Force criteria [21]), frontotemporal dementia (FTD, according to Manchester/Neary criteria [22]), unspecified dementia (where specific dementia diagnosis is not ascertained; here DUNS) and other dementia subtypes (grouping miscellaneous dementia disorders such as corticobasal degeneration or alcohol related dementias; here OD [16]).
The third source was the Swedish mortality registry, which is maintained by the Swedish National Board of Health and Welfare. Here records are maintained of all deaths in Sweden. The physician issuing the death certificate used the International Statistical Classification of Diseases and Related Health Problems [23] as the standard diagnostic tool.
The three data sources were merged using the Swedish 10-digit unique national identification number. In the main analysis, inclusion criteria were having at least one CSF GFAP result with a MMSE score registered within 24 months before, or 36 months after sampling. MMSE is a well-known measure of tracking cognitive decline; a scoring system ranging from 0 to 30 points, with lower scores indicating greater cognitive impairment, as described in [24]. In this study, MMSE was used as a proxy for disease severity, as done previously [25, 26]. When multiple CSF analyses were available for the same individual, the measurement closest to the date of diagnosis was used. In the resulting dataset, a subset of individuals included data on Aβ42 and Aβ40 concentrations, where we performed a subgroup analysis.
Cerebrospinal Fluid Measurements
All CSF analyses were performed in clinical practice by board-certified laboratory technicians using procedures accredited by the Swedish Board for Accreditation and Conformity Assessment (SWEDAC). CSF GFAP concentration was measured using an in-house enzyme-linked immunosorbent assay (ELISA), as previously described [27]. At the start of the study, Aβ1-42 and Aβ1-40 concentrations were measured by MesoScale Discovery (MSD)™. Starting in November 2019, both Aβ1-42 and Aβ1-40 were measured by Lumipulse chemiluminescent enzyme immunoassay (CLEIA)™. Sarstedt polypropylene tubes were used, and the CSF was centrifuged to count cells before being sent to the neurochemistry laboratory. During overnight transport CSF samples were transported on dry ice. Additional information on the assays and their harmonization, CSF collection, processing and storage procedures has been described previously [28, 29].
Cut-offs for 10xAβ1-42/Aβ1-40 were >0.89 on the MSD platform and changed to >0.61 when Lumipulse was introduced in 2019. One year after the platform change, the cut-off was reassessed and changed to >0.72 based on the bimodal distribution observed when more than 2,700 samples were analyzed in a routine hospital laboratory setting [28]. The reference values for normal biomarker concentrations for GFAP are <750 pg/mL in ages 20–60 years, and <1,250 pg/mL in ages 60 years and older. The cut-offs were developed and validated locally at the neurochemistry laboratory in Gothenburg [27].
Statistics
The Shapiro-Wilk method was used to test for normal distributions in the variables of interest and log10-transformed if significantly skewed. All CSF biomarkers and MMSE scores were significantly skewed (all p > 0.05), thus log10-transformed for the subsequent analyses. Mainly three statistical analyses were used in this study: Analysis of covariance (ANCOVA), receiver operating characteristic (ROC) analysis and multiple linear regression.
ANCOVA was performed to investigate levels of GFAP and MMSE, first on the complete material, and then across diagnoses, using Šidák correction for multiple testing. Age and sex were included as covariates. Effect sizes were measured by partial eta squared (η2). The groups differing significantly from each other in GFAP levels were then tested with ROC analysis, to evaluate CSF GFAP as a potential differential diagnostic tool. Lastly, in the subgroup of individuals with Aβ42/Aβ40 (Aβ) data, we used time-derived cut-offs, to stratify according to Aβ+ and Aβ− status. Then, we used ANCOVA to test for differences in GFAP concentrations depending on Aβ status, in three scenarios: (1) all subjects; (2) only the EAD and LAD groups; (3) all groups except EAD and LAD.
In the multiple linear regressions performed, all included GFAP, age and sex as independent variables. First, models were fitted against Aβ42, the Aβ42/40 ratio and MMSE at baseline, and then to time of sampling and number of years to death. Adjusted R2 values were used to assess model fit. B was the unadjusted coefficient and represented the direct effect of GFAP on the outcome variable, all other predictors held constant. β was the adjusted coefficient, representing the number of standard deviations the dependent variable changed, when all other variables were held constant, thus the relative strength of GFAP relative to the other predictors. All analyses were adjusted for age and sex.
To see if the selected cohort, based on GFAP measurements and MMSE scores, differed from the general dementia population in terms of age, sex, MMSE and CSF biomarkers and, thus, could have inherited a systematic error, a bias-analysis was conducted. Additional information can be found under Bias-analysis in the online supplementary material (for all online suppl. material, see https://doi.org/10.1159/000550601). Statistical significance was determined at p < 0.05 and was two-tailed. Statistics and tables were computed using SPSS version 29.0.2 (IBM, New York). The figures were made using the “ggplot2” package in R version 4.4.1 (R foundation for statistical computing, Vienna, Austria).
Results
Cohort Characteristics
In total, 1,345 (mean age [SD], 72 [8.0] years, 49% female) individuals were included (Table 1). Of these, in a subgroup containing Aβ concentrations, 197 individuals had data on Aβ42, and 184 individuals included data on both Aβ42 and Aβ40, for additional information see Table in the online supplementary material. 541 individuals included data on date of death. In the whole cohort, there was a significant impact of diagnosis on age (p < 0.001). EAD and FTD patients were younger than the other diagnostic groups (p < 0.01). Individuals with OD were older than EAD and FTD, but younger than the other diagnostic categories (all p < 0.05), albeit not significantly different from LBD and PDD. LBD and PDD did not differ in age (p > 0.05). The sex distribution ranged from 16.9% to 18.5% females in LBD and PDD, respectively, to 61.1% in LAD, see Table 1. Ages were similar in the subgroup with CSF Aβ concentrations, while the sex distributions differed slightly, as seen in online supplementary Table. In the whole cohort, there was a significant, but with low effect size, impact of diagnosis on MMSE score (p = 0.012, η2 = 0.15). Higher MMSE scores were seen in FTD compared to EAD (mean difference [MD] = 1.2, standard error [SE] = 1.0, p = 0.005). In the whole cohort, no sex-related differences in GFAP concentrations were seen (all p > 0.05). No bias was detected for the GFAP-group or any of the investigated variables, compared to individuals without GFAP measurements; see Bias-analysis in the online supplementary material.
Table 1.
Cohort characteristics
| | Total (% F) | Age, years | GFAP | MMSE | RD | Survival, yearsa |
|---|---|---|---|---|---|---|
| Total | 1,345 (49) | | | | 541 | |
| Mean (SD) | | 72 (8.0) | 927 (2,965) | 22 (5) | | 4.1 (2.4) |
| Mdn (IQR) | | 72 (67–77) | 730 (520–980) | 23 (19–26) | | 4.0 (2.1–5.8) |
| EAD | 191 (58) | | | | 47 | |
| Mean (SD) | | 62 (5.0) | 654 (295) | 21 (5) | | 5.3 (2.1) |
| Mdn (IQR) | | 63 (59–65) | 630 (460–830) | 22 (18–25) | | 5.1 (3.7–7.1) |
| LAD | 388 (61) | | | | 143 | |
| Mean (SD) | | 75 (5.3) | 951 (1,472) | 22 (5) | | 4.3 (2.5) |
| Mdn (IQR) | | 74 (70–78) | 820 (620–1,045) | 23 (19–26) | | 4.4 (2.2–6.3) |
| VaD | 146 (38) | | | | 67 | |
| Mean (SD) | | 74 (7.1) | 1,690 (8,584) | 22 (4) | | 3.6 (2.3) |
| Mdn (IQR) | | 75 (69–79) | 790 (530–1,180) | 22 (20–25) | | 3.5 (1.9–5.3) |
| MIX | 151 (53) | | | | 68 | |
| Mean (SD) | | 75 (5.9) | 852 (465) | 22 (5) | | 4.0 (2.7) |
| Mdn (IQR) | | 75 (71–79) | 770 (590–1,000) | 23 (19–26) | | 4.1 (1.5–6.0) |
| FTD | 71 (39) | | | | 32 | |
| Mean (SD) | | 65 (8.9) | 693 (351) | 24 (5) | | 3.9 (2.2) |
| Mdn (IQR) | | 66 (61–72) | 590 (460–880) | 25 (22–28) | | 4.0 (1.9–5.1) |
| LBD | 65 (17) | | | | 32 | |
| Mean (SD) | | 72 (6.7) | 774 (355) | 23 (5) | | 4.1 (2.5) |
| Mdn (IQR) | | 73 (69–76) | 730 (530–940) | 23 (21–26) | | 4.2 (1.8–5.5) |
| PDD | 27 (19) | | | | 14 | |
| Mean (SD) | | 71 (8.4) | 583 (364) | 22 (4) | | 3.5 (1.7) |
| Mdn (IQR) | | 73 (65–77) | 550 (330–720) | 23 (20–25) | | 3.6 (1.8–5.0) |
| OD | 55 (31) | | | | 24 | |
| Mean (SD) | | 70 (8.3) | 888 (1,027) | 21 (5) | | 3.3 (2.2) |
| Mdn (IQR) | | 71 (66–75) | 680 (460–1,030) | 21 (18–25) | | 2.5 (1.8–5.3) |
| DUNS | 251 (45) | | | | 114 | |
| Mean (SD) | | 73 (7.9) | 870 (830) | 22 (4) | | 4.0 (2.4) |
| Mdn (IQR) | | 74 (68–79) | 710 (500–1,000) | 23 (19–25) | | 3.7 (2.2–5.5) |
F, female; IQR, interquartile range; Mdn, median; MMSE, Mini-Mental State Examination; RD, registered deaths.
aTime from sampling to death.
GFAP across Diagnoses
Figure 1 shows the unadjusted distribution of GFAP across all groups. GFAP concentrations were log10-transformed for subsequent analysis. In the whole cohort, there was a significant, but with low effect size, impact of diagnosis on GFAP concentrations (p = 0.015, η2 = 0.014). PDD had lower concentrations of GFAP than EAD ([MD] = −0.17, SE = 0.05, p = 0.025), LAD (MD = −0.18, SE = 0.05, p = 0.006), and VaD (MD = −0.20, SE = 0.05, p = 0.004). No differences in mean concentrations were observed between EAD and LAD (MD = 0.01 SE = 0.03, p = 1.0), PDD and LBD (MD = −0.15, SE = 0.06, p = 0.176), or in any of the other groups (all p > 0.05).
Fig. 1.
Distributions of CSF GFAP across diagnostic groups in SveDem.
Subsequently, ROC curves were produced. The performance for GFAP in distinguishing PDD from other dementias in falling order was PDD to LAD (AUROC = 0.74, CI = 0.64–0.84, p < 0.001), VaD (AUROC = 0.71, CI =0.61–0.81, p < 0.001), and EAD (AUROC = 0.59, CI = 0.47–0.71, p = 0.13).
In the subgroup analysis based on Aβ42/Aβ40 status, GFAP concentrations did not differ in any of the three scenarios (all p > 0.05). Visualizations for all and non-AD individuals are shown in Figure 2. The number of Aβ-negative individuals in the analysis including only AD patients (n = 5) was too small for a meaningful comparison and was, therefore, excluded from the visualizations.
Fig. 2.
a, b Distributions of CSF GFAP across diagnostic groups based on Aβ status.
Biomarker Associations
We first investigated the ability of GFAP to predict disease severity using MMSE scores at baseline as a proxy. In the whole cohort, higher GFAP yielded lower MMSE (B = −0.028, SE = 0.013, β = −0.062, R2 = 0.002, p = 0.032), and higher Aβ42 (B = 0.192, SE = 0.055, β = 0.26, R2 = 0.051, p < 0.001). When stratifying by diagnosis, GFAP associated with decreasing MMSE in MIX (B = −0.092, SE = 0.041, β = −0.19, R2 = 0.014, p = 0.027), but not in any other groups (all p > 0.05).
GFAP predicted changes in Aβ42 concentrations in DUNS (B = 0.409, SE = 0.19, β = 0.60, R2 = 0.12, p = 0.037), and in LAD (B = 0.16, SE = 0.072, β = 0.25, R2 = 0.040, p = 0.033), but not in any other group (all p > 0.05). In the subgroup analysis, comparing GFAP concentrations according to Aβ status in three scenarios, we did not find any differences in any of the three scenarios (all p > 0.05). GFAP did not significantly predict or correlate with Aβ42/40 concentrations, nor predict the time from sampling to death.
Discussion
We tested the diagnostic and prognostic abilities of GFAP in a large dataset of patients containing a wide range of dementia diagnoses. Altogether, GFAP did not show clinically relevant discriminative abilities, nor associations, in any of the investigated dementias.
We found lower levels of GFAP in the PDD group compared with EAD, LAD, and VaD. This finding is difficult to interpret; as to our knowledge, no previous studies have compared CSF GFAP levels between different dementia groups, limiting us from a direct comparison to previous literature. Evidence pointing toward higher levels of CSF GFAP in PD has been shown by Li et al. [30] where the authors observed a positive association of the marker with enlarged perivascular spaces on MRI and to cognitive decline in certain brain regions in Parkinson’s patients. The marker has also showed promise as a prognostic marker in PDD in a longitudinal study led by Liu et al. [13], where the authors investigated CSF GFAP as a predictive biomarker for conversion to dementia. Here, high levels of baseline GFAP correlated with a faster cognitive decline and predicted conversion to dementia. In the present study, however, when investigating associations of CSF GFAP levels in relation to the variable time to death, no significant associations to this variable were seen in any of the groups. However, our study was cross-sectional, limiting us from making direct comparisons.
We further tested the performance of GFAP as a tool in dementia differential diagnostics, where the marker performed poorly to moderately; at a level unlikely to be sufficient for clinical use. Interestingly, even though similar sample sizes and distributions were seen in LBD and PDD, no differences in GFAP concentrations were found when comparing the two groups, suggesting similar neuroinflammatory levels between diagnoses and supporting the findings of Schulz et al. [31]. However, this study was neither designed nor powered to detect differences or similarities between LBD and PDD.
In the full sample, increasing GFAP concentrations associated with higher MMSE scores and Aβ42 levels. Stratified, CSF GFAP showed a weak contribution to MMSE prediction in MIX (R2 = 0.034), but not in other diagnostic groups. In the literature there is evidence of more severe cognitive impairment in MIX patients compared to “pure” AD or VaD [32], which could be related to the signal we see in this study.
GFAP predicted Aβ42 levels in LAD and DUNS, suggesting associations between neuroinflammatory changes and amyloid burden in these groups. Both being two of the older groups and having some of the higher amyloid plaque burdens, these factors might have given rise to the general neuroinflammation we observed. LAD and DUNS were also the largest categories in the study population, which provided greater power to detect these associations. No previous studies have directly explored the association between CSF Aβ42 and CSF GFAP in LAD, however, positive associations with Aβ PET, in groups similar to LAD have been found in a study by Pereira et al. [33] corroborating our finding and suggesting that reactive astrogliosis occurs concurrent with amyloid pathology.
However, when testing associations between CSF GFAP and the AD-specific Aβ42/Aβ40 ratio [34], no significant associations were found in the AD groups, nor in any of the other investigated diagnoses. Neither did we observe any significant differences when comparing CSF GFAP levels between all and non-AD groups, respectively, in terms of Aβ status. These results corroborate the findings of Benédet et al. [11] where they did not find meaningful differences in CSF GFAP levels between any of the AD continuum groups, and showed that plasma GFAP is more suitable for detecting Aβ pathology than its CSF counterpart.
Reasons for the discrepancy in performance seen between CSF and plasma GFAP might be attributed to biofluid differences and pre-analytical factors. CSF GFAP might be a better reflection of reactive astrocytes in response to general neuroinflammatory changes, while plasma GFAP is more closely associated with reactive astrogliosis due to Aβ burden [11, 35].
Further, previous studies have shown that CSF GFAP is freeze-thaw cycle (FTC) sensitive, with marked decreases across cycles, while serum GFAP remained unaffected [35, 36]. Also, the CSF GFAP levels might be assay-specific, with ELISA consistently underestimating concentrations compared to other platforms [37]. Also, a recent study by Evertsson et al. [38] found that CSF GFAP is affected by several different factors related to transportation and the storage of the sample. Measures could relatively easily be taken to account for these; a small, filled and sealed tube at time of collection (microtube), improves sample stability for up to 3 weeks. Since no information concerning freeze-thaw cycles was included in the dataset, and a lack of pre-analytical handling for specific samples in our cohort, these caveats could not be adjusted for. Future studies should strive to adjust for the above factors to enable harmonization of results.
This study had limitations. We did not have access to a control group with healthy subjects, nor were data on APOE4 status present. We did not have information on the ethnicity of our subjects. In PDD, the sample size was small, and the patients were predominantly male. Also, the diagnostic criteria used for AD, LBD and FTD were based on previous standards. As we did not have access to the patients’ full medical records, we could not retrospectively apply current diagnostic guidelines. Further, using MMSE as a proxy of disease severity in non-AD cases, particularly FTD, has shown acceptable performance as a global severity marker; however, its use is not recommended in early FTD stages [39, 40]. Altogether, these aspects could impact the generalizability of our results. However, the main strengths of this study were the large sample size, the broad referral base and the wide range of dementia diagnoses represented in a non-selected real-world setting.
In conclusion, we did not see a clinically relevant diagnostic or prognostic performance of CSF GFAP in dementia in this study; thus, we do not currently recommend its use in clinical practice. Importantly, these findings could save valuable time and resources in the healthcare system. Finally, the marker showed promise in distinguishing PDD from other dementia groups, and we, thus, encourage further investigation of the marker in larger studies.
Statement of Ethics
This study was approved by the Swedish Research Ethics Review Authority (2021/03304). Patient written consent was not required in accordance with national guidelines; however, each patient has the right to obtain a copy of the information that is registered if requested. Each patient must be informed about the registration and has a right to decline participation. The patient has the right to have their data removed from the registry. An ethical approval from a regional Ethics Committee for each research project where SveDem data will be used is needed.
Conflict of Interest Statement
P.E., T.B.S., U.A., and I.S. declared no conflicts of interest. K.B. has served as a consultant and at advisory boards for AbbVie, AC Immune, ALZPath, AriBio, Beckman-Coulter, BioArctic, Biogen, Eisai, Lilly, Moleac Pte. Ltd, Neurimmune, Novartis, Ono Pharma, Prothena, Quanterix, Roche Diagnostics, Sanofi, and Siemens Healthineers; has served at data monitoring committees for Julius Clinical and Novartis; has given lectures, produced educational materials, and participated in educational programs for AC Immune, Biogen, Celdara Medical, Eisai, and Roche Diagnostics; and is a co-founder of Brain Biomarker Solutions in Gothenburg AB (BBS), which is a part of the GU Ventures Incubator Program, outside the work presented in this paper. S.K. has served at scientific advisory boards, speaker, and/or as consultant for Roche, Eli Lilly, Geras Solutions, Optoceutics, Biogen, Eisai, Merry Life, Triolab, Novo Nordisk, and Bioarctic, unrelated to present study content. M.E. has served as consultant for Bioarctic AB, Roche, Eli Lilly, Biogen/Eisai, and Novo Nordisk and given lectures in symposia sponsored by Roche. H.Z. has served at scientific advisory boards and/or as a consultant for AbbVie, Acumen, Alector, Alzinova, ALZpath, Amylyx, Annexon, Apellis, Artery Therapeutics, AZTherapies, Cognito Therapeutics, CogRx, Denali, Eisai, LabCorp, Merry Life, NervGen, Novo Nordisk, Optoceutics, Passage Bio, Pinteon Therapeutics, Prothena, Quanterix, Red Abbey Labs, reMYND, Roche, Samumed, Siemens Healthineers, Triplet Therapeutics, and Wave; has given lectures sponsored by Alzecure, BioArctic, Biogen, Cellectricon, Fujirebio, Lilly, Novo Nordisk, Roche, and WebMD; and is a co-founder of Brain Biomarker Solutions in Gothenburg AB (BBS), which is a part of the GU Ventures Incubator Program (outside submitted work). L.J. has received research funding (unrelated to the study) from Novo Nordisk and consulting fees from H. Lundbeck, TEVA, Eli Lilly, and Novo Nordisk.
Funding Sources
P.E. was supported by the scholarship “Testa Forskning” by Göteborgs Läkaresällskap (GLS). K.B. is supported by the Swedish Research Council (#2017-00915 and #2022-00732); the Swedish Alzheimer Foundation (#AF-930351, #AF-939721, #AF-968270, and #AF-994551); Hjärnfonden, Sweden (#ALZ2022-0006 and #FO2024-0048-TK-130); the Swedish state under the agreement between the Swedish government and the County Councils, the ALF-agreement (#ALFGBG-965240 and #ALFGBG-1006418); the European Union Joint Program for Neurodegenerative Disorders (JPND2019-466-236), the Alzheimer’s Association 2021 Zenith Award (ZEN-21-848495); the Alzheimer’s Association 2022-2025 Grant (SG-23-1038904 QC); La Fondation Recherche Alzheimer (FRA), Paris, France; the Kirsten and Freddy Johansen Foundation, Copenhagen, Denmark; and Familjen Rönströms Stiftelse, Stockholm, Sweden. S.K. was financed by grants from the Swedish state under the agreement between the Swedish government and the county councils the ALF-agreement (ALFGBG-1005471, ALFGBG-965923, ALFGBG-81392, ALF GBG-771071); the Alzheimerfonden (AF-842471, AF-737641, AF-929959, AF-939825); the Swedish Research Council (2019-02075, 2019-02075_15); stiftelsen Psykiatriska Forskningsfonden; and the Swedish Brain Foundation FO2024-0097. M.E. was supported by grants from Vinnova (#2021-02680); the Swedish research council grant (#2020-02014); Johanniterorden; and the Brain Foundation and the regional agreement on medical training and clinical research between the Stockholm County council and the Karolinska Institutet (ALF). I.S. was financed by grants from the Swedish state under the agreement between the Swedish government and the county councils, the ALF-agreement (ALFGBG-1006423); the Swedish Research Council (2017-00639, 2022-00882); and the Kamprad Family Foundation for Entrepreneurship, Research & Charity. H.Z. is a Wallenberg Scholar and a Distinguished Professor at the Swedish Research Council supported by grants from the Swedish Research Council (#2023-00356, #2022-01018 and #2019-02397); the European Union’s Horizon Europe research and innovation programme under grant agreement No 101053962; Swedish State Support for Clinical Research (#ALFGBG-71320); the Alzheimer Drug Discovery Foundation (ADDF), USA (#201809-2016862); the AD Strategic Fund and the Alzheimer’s Association (#ADSF-21-831376-C, #ADSF-21-831381-C, #ADSF-21-831377-C, and #ADSF-24-1284328-C); the European Partnership on Metrology, co-financed from the European Union’s Horizon Europe Research and Innovation Programme and by the Participating States (NEuroBioStand, #22HLT07); the Bluefield Project; Cure Alzheimer’s Fund; the Olav Thon Foundation; the Erling-Persson Family Foundation; Familjen Rönströms Stiftelse; Stiftelsen för Gamla Tjänarinnor, Hjärnfonden Sweden (#FO2022-0270); the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 860197 (MIRIADE); the European Union Joint Programme – Neurodegenerative Disease Research (JPND2021-00694); the National Institute for Health and Care Research University College London Hospitals Biomedical Research Centre; the Dementia Research Institute at UCL (UKDRI-1003); and an anonymous donor. L.J. has received research funding from European Commission (Horizon Europe); FORTE – Swedish Research Council for Health, Working Life, and Welfare; and Vinnova – Swedish Innovation Agency.
Author Contributions
P.E. conceived and designed the study, performed the statistical analysis, interpreted the results, drafted and revised the manuscript. T.B.S. supervised, designed, analyzed and interpreted the results, and aided in the drafting of the manuscript. S.K. interpreted the data and critically reviewed the manuscript for important intellectual content. I.S. interpreted the data and critically reviewed the manuscript for important intellectual content. L.J. aided in statistical analysis and critically reviewed the manuscript for important intellectual content. U.A. supervised, aided in the statistical analysis, and critically reviewed the manuscript for important intellectual content. K.B. aided in the interpretation of data and critically reviewed the manuscript for important intellectual content. M.E. aided the acquisition of data and critically reviewed the manuscript for important intellectual content. H.Z. supervised, aided in the design of the study and statistical analysis, and reviewed the manuscript for important intellectual content. Drs. P.E., T.B.S., and H.Z. had full access to all the data in the study and take responsibility for the integrity of the data and accuracy of the data analysis.
Funding Statement
P.E. was supported by the scholarship “Testa Forskning” by Göteborgs Läkaresällskap (GLS). K.B. is supported by the Swedish Research Council (#2017-00915 and #2022-00732); the Swedish Alzheimer Foundation (#AF-930351, #AF-939721, #AF-968270, and #AF-994551); Hjärnfonden, Sweden (#ALZ2022-0006 and #FO2024-0048-TK-130); the Swedish state under the agreement between the Swedish government and the County Councils, the ALF-agreement (#ALFGBG-965240 and #ALFGBG-1006418); the European Union Joint Program for Neurodegenerative Disorders (JPND2019-466-236), the Alzheimer’s Association 2021 Zenith Award (ZEN-21-848495); the Alzheimer’s Association 2022-2025 Grant (SG-23-1038904 QC); La Fondation Recherche Alzheimer (FRA), Paris, France; the Kirsten and Freddy Johansen Foundation, Copenhagen, Denmark; and Familjen Rönströms Stiftelse, Stockholm, Sweden. S.K. was financed by grants from the Swedish state under the agreement between the Swedish government and the county councils the ALF-agreement (ALFGBG-1005471, ALFGBG-965923, ALFGBG-81392, ALF GBG-771071); the Alzheimerfonden (AF-842471, AF-737641, AF-929959, AF-939825); the Swedish Research Council (2019-02075, 2019-02075_15); stiftelsen Psykiatriska Forskningsfonden; and the Swedish Brain Foundation FO2024-0097. M.E. was supported by grants from Vinnova (#2021-02680); the Swedish research council grant (#2020-02014); Johanniterorden; and the Brain Foundation and the regional agreement on medical training and clinical research between the Stockholm County council and the Karolinska Institutet (ALF). I.S. was financed by grants from the Swedish state under the agreement between the Swedish government and the county councils, the ALF-agreement (ALFGBG-1006423); the Swedish Research Council (2017-00639, 2022-00882); and the Kamprad Family Foundation for Entrepreneurship, Research & Charity. H.Z. is a Wallenberg Scholar and a Distinguished Professor at the Swedish Research Council supported by grants from the Swedish Research Council (#2023-00356, #2022-01018 and #2019-02397); the European Union’s Horizon Europe research and innovation programme under grant agreement No 101053962; Swedish State Support for Clinical Research (#ALFGBG-71320); the Alzheimer Drug Discovery Foundation (ADDF), USA (#201809-2016862); the AD Strategic Fund and the Alzheimer’s Association (#ADSF-21-831376-C, #ADSF-21-831381-C, #ADSF-21-831377-C, and #ADSF-24-1284328-C); the European Partnership on Metrology, co-financed from the European Union’s Horizon Europe Research and Innovation Programme and by the Participating States (NEuroBioStand, #22HLT07); the Bluefield Project; Cure Alzheimer’s Fund; the Olav Thon Foundation; the Erling-Persson Family Foundation; Familjen Rönströms Stiftelse; Stiftelsen för Gamla Tjänarinnor, Hjärnfonden Sweden (#FO2022-0270); the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 860197 (MIRIADE); the European Union Joint Programme – Neurodegenerative Disease Research (JPND2021-00694); the National Institute for Health and Care Research University College London Hospitals Biomedical Research Centre; the Dementia Research Institute at UCL (UKDRI-1003); and an anonymous donor. L.J. has received research funding from European Commission (Horizon Europe); FORTE – Swedish Research Council for Health, Working Life, and Welfare; and Vinnova – Swedish Innovation Agency.
Data Availability Statement
The data from this study are not publicly available due to information that could compromise the privacy of research participants. The data could be shared upon reasonable request. Further inquiries can be directed to the corresponding author.
Supplementary Material.
References
- 1. Inoue Y, Shue F, Bu G, Kanekiyo T. Pathophysiology and probable etiology of cerebral small vessel disease in vascular dementia and Alzheimer’s disease. Mol Neurodegener. 2023;18(1):46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Jack CR Jr, Knopman DS, Jagust WJ, Petersen RC, Weiner MW, Aisen PS, et al. Tracking pathophysiological processes in Alzheimer’s disease: an updated hypothetical model of dynamic biomarkers. Lancet Neurol. 2013;12(2):207–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Bartl M, Dakna M, Galasko D, Hutten SJ, Foroud T, Quan M, et al. Biomarkers of neurodegeneration and glial activation validated in Alzheimer’s disease assessed in longitudinal cerebrospinal fluid samples of Parkinson’s disease. PLoS One. 2021;16(10):e0257372. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Smirnov DS, Galasko D, Edland SD, Filoteo JV, Hansen LA, Salmon DP. Cognitive decline profiles differ in Parkinson disease dementia and dementia with Lewy bodies. Neurology. 2020;94(20):e2076–87. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Chin KS, Holper S, Loveland P, Churilov L, Yassi N, Watson R. Prevalence of cerebral microbleeds in Alzheimer’s disease, dementia with Lewy bodies and Parkinson’s disease dementia: a systematic review and meta-analysis. Neurobiol Aging. 2024;134:74–83. [DOI] [PubMed] [Google Scholar]
- 6. Jack CR Jr, Andrews JS, Beach TG, Buracchio T, Dunn B, Graf A, et al. Revised criteria for diagnosis and staging of Alzheimer’s disease: Alzheimer’s Association Workgroup. Alzheimers Dement. 2024;20(8):5143–69. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Wilson DM 3rd, Cookson MR, Van Den Bosch L, Zetterberg H, Holtzman DM, Dewachter I. Hallmarks of neurodegenerative diseases. Cell. 2023;186(4):693–714. [DOI] [PubMed] [Google Scholar]
- 8. Kumar A, Fontana IC, Nordberg A. Reactive astrogliosis: a friend or foe in the pathogenesis of Alzheimer’s disease. J Neurochem. 2023;164(3):309–24. [DOI] [PubMed] [Google Scholar]
- 9. Bellaver B, Ferrari-Souza JP, Uglione da Ros L, Carter SF, Rodriguez-Vieitez E, Nordberg A, et al. Astrocyte biomarkers in alzheimer disease. Neurology. 2021;96(24):e2944–55. [DOI] [PubMed] [Google Scholar]
- 10. Yang Z, Wang KK. Glial fibrillary acidic protein: from intermediate filament assembly and gliosis to neurobiomarker. Trends Neurosci. 2015;38(6):364–74. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Benedet AL, Milà-Alomà M, Vrillon A, Ashton NJ, Pascoal TA, Lussier F, et al. Differences between plasma and cerebrospinal fluid glial fibrillary acidic protein levels across the alzheimer disease continuum. JAMA Neurol. 2021;78(12):1471–83. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Ishiki A, Kamada M, Kawamura Y, Terao C, Shimoda F, Tomita N, et al. Glial fibrillar acidic protein in the cerebrospinal fluid of Alzheimer’s disease, dementia with Lewy bodies, and frontotemporal lobar degeneration. J Neurochem. 2016;136(2):258–61. [DOI] [PubMed] [Google Scholar]
- 13. Liu T, Zuo H, Ma D, Song D, Zhao Y, Cheng O. Cerebrospinal fluid GFAP is a predictive biomarker for conversion to dementia and Alzheimer’s disease-associated biomarkers alterations among de novo Parkinson’s disease patients: a prospective cohort study. J Neuroinflammation. 2023;20(1):167. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Tang Y, Han L, Li S, Hu T, Xu Z, Fan Y, et al. Plasma GFAP in Parkinson's disease with cognitive impairment and its potential to predict conversion to dementia. NPJ Parkinsons Dis. 2023;9(1):23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Korley FK, Jain S, Sun X, Puccio AM, Yue JK, Gardner RC, et al. Prognostic value of day-of-injury plasma GFAP and UCH-L1 concentrations for predicting functional recovery after traumatic brain injury in patients from the US TRACK-TBI cohort: an observational cohort study. Lancet Neurol. 2022;21(9):803–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Religa D, Fereshtehnejad SM, Cermakova P, Edlund AK, Garcia-Ptacek S, Granqvist N, et al. SveDem, the Swedish Dementia Registry - a tool for improving the quality of diagnostics, treatment and care of dementia patients in clinical practice. PLoS One. 2015;10(2):e0116538. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.SveDem. Årsrapport 2021. 2021.
- 18. American Psychological Association . Diagnostic and statistical manual of mental disorders (DSM–5). 2013. [DOI] [PubMed] [Google Scholar]
- 19. World Health Organization . The ICD-10 classification of mental and behavioural disorders: diagnostic criteria for research. Geneva: World Health Organization; 1993. p. 248. [Google Scholar]
- 20. McKeith IG. Consensus guidelines for the clinical and pathologic diagnosis of dementia with Lewy bodies (DLB): report of the Consortium on DLB International Workshop. J Alzheimers Dis. 2006;9(3 Suppl l):417–23. [DOI] [PubMed] [Google Scholar]
- 21. Emre M, Aarsland D, Brown R, Burn DJ, Duyckaerts C, Mizuno Y, et al. Clinical diagnostic criteria for dementia associated with Parkinson’s disease. Mov Disord. 2007;22(12):1689–837. [DOI] [PubMed] [Google Scholar]
- 22. Clinical and neuropathological criteria for frontotemporal dementia. The Lund and Manchester Groups. J Neurol Neurosurg Psychiatry. 1994;57(4):416–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. World Health Organization . International statistical classification of diseases and related health problems (ICD). 2024. [Available from: https://www.who.int/standards/classifications/classification-of-diseases
- 24. Creavin ST, Wisniewski S, Noel-Storr AH, Trevelyan CM, Hampton T, Rayment D, et al. Mini-Mental State Examination (MMSE) for the detection of dementia in clinically unevaluated people aged 65 and over in community and primary care populations. Cochrane Database Syst Rev. 2016;2016(1):Cd011145. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Kennedy RE, Cutter GR, Wang G, Schneider LS. Using baseline cognitive severity for enriching Alzheimer’s disease clinical trials: how does Mini-Mental State Examination predict rate of change? Alzheimers Dement. 2015;1(1):46–52. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Mueller J, Cammermeyer G. Patient-specific cognitive profiles in the detection of dementia subtypes: a proposal. Alzheimers Dement. 2023;19(10):4743–52. [DOI] [PubMed] [Google Scholar]
- 27. Rosengren LE, Wikkelsø C, Hagberg L. A sensitive ELISA for glial fibrillary acidic protein: application in CSF of adults. J Neurosci Methods. 1994;51(2):197–204. [DOI] [PubMed] [Google Scholar]
- 28. Gobom J, Parnetti L, Rosa-Neto P, Vyhnalek M, Gauthier S, Cataldi S, et al. Validation of the LUMIPULSE automated immunoassay for the measurement of core AD biomarkers in cerebrospinal fluid. Clin Chem Lab Med. 2022;60(2):207–19. [DOI] [PubMed] [Google Scholar]
- 29. Palmqvist S, Zetterberg H, Blennow K, Vestberg S, Andreasson U, Brooks DJ, et al. Accuracy of brain amyloid detection in clinical practice using cerebrospinal fluid β-Amyloid 42: a cross-validation Study against amyloid positron emission tomography. JAMA Neurol. 2014;71(10):1282–9. [DOI] [PubMed] [Google Scholar]
- 30. Li Q, Bai X, Zhang S, Cheng Y, Qian A, Chen S, et al. The Association of enlarged perivascular space with cerebrospinal fluid GFAP in early Parkinson’s disease. Acad Radiol. 2025;33(2):556–63. [DOI] [PubMed] [Google Scholar]
- 31. Schulz I, Kruse N, Gera RG, Kremer T, Cedarbaum J, Barbour R, et al. Systematic assessment of 10 biomarker candidates focusing on α-Synuclein-Related disorders. Mov Disord. 2021;36(12):2874–87. [DOI] [PubMed] [Google Scholar]
- 32. Sarhan M, Wohlfeld C, Perry-Mills A, Meyers J, Fadel J, Murphy EA, et al. The pathophysiology of mixed Alzheimer’s disease and vascular dementia. Theranostics. 2025;15(18):9793–818. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Pereira JB, Janelidze S, Smith R, Mattsson-Carlgren N, Palmqvist S, Teunissen CE, et al. Plasma GFAP is an early marker of amyloid-β but not tau pathology in Alzheimer’s disease. Brain. 2021;144(11):3505–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Hansson O, Lehmann S, Otto M, Zetterberg H, Lewczuk P. Advantages and disadvantages of the use of the CSF Amyloid β (Aβ) 42/40 ratio in the diagnosis of Alzheimer’s Disease. Alzheimers Res Ther. 2019;11(1):34. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Abdelhak A, Hottenrott T, Morenas-Rodríguez E, Suárez-Calvet M, Zettl UK, Haass C, et al. Glial activation markers in CSF and serum from patients with primary progressive multiple sclerosis: potential of serum GFAP as disease severity marker? Front Neurol. 2019;10:280. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Simrén J, Weninger H, Brum WS, Khalil S, Benedet AL, Blennow K, et al. Differences between blood and cerebrospinal fluid glial fibrillary Acidic protein levels: the effect of sample stability. Alzheimers Dement. 2022;18(10):1988–92. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Arslan B, Rembeza E, Rasch S, Andreasson U, Blennow K, Zetterberg H, et al. Method comparison of plasma and CSF GFAP immunoassays across multiple platforms. Clin Chem Lab Med. 2025;63(12):2447–59. [DOI] [PubMed] [Google Scholar]
- 38. Evertsson B, Remahl IN, Hietala MA, Finn A. Glial fibrillary acidic protein in cerebrospinal fluid in humans is sensitive to various pre-analytical conditions: possible explanation and solution. Front Neurol. 2025;16:1627405–2025. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. de Boer L, Poos JM, Van Den Berg E, De Houwer JFH, Swartenbroekx T, Dopper EGP, et al. Montreal cognitive assessment vs the mini-mental State examination as a screening tool for patients with genetic frontotemporal dementia. Neurology. 2025;104(5):e213401. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Ruiz-Garcia R, Yu S, Richardson L, Roberts A, Pasternak S, Stewart C, et al. Comparison of behavior-related features in the MMSE sentence in behavioral variant frontotemporal dementia and alzheimer’s disease. Front Aging Neurosci. 2021;13:733153. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data from this study are not publicly available due to information that could compromise the privacy of research participants. The data could be shared upon reasonable request. Further inquiries can be directed to the corresponding author.
