Pooling Alzheimer's disease clinical trial data to develop personalized medicine approaches is easier said than done: A proof-of-principle study and call to action.

With the advent of the first generation of disease-modifying treatments for Alzheimer's disease, it is clearer now more than ever that the field needs to move toward personalized medicine. Pooling data from past trials may help identify subgroups most likely to benefit from specific treatments and thus inform future trial design. In this perspective, we report on our effort to pool data from past Alzheimer's disease trials to identify patients most likely to respond to different treatments. We delineate challenges and hurdles, from our proof-of-principle study, for which we requested access to trial datasets from various pharmaceutical companies and encountered obstacles in the process of arranging data-sharing agreements through legal departments. Six phase I-III trials from three sponsors provided access to their data (total n = 3170), which included demographic information, vital signs, primary and secondary endpoints, and in a small subset, cerebrospinal fluid amyloid (n = 165, 5.2%) and tau (n = 212, 6.7%). Data could be analyzed only within specific data access platforms, limiting potential harmonization with data provided through other platforms. Limited overlap in terms of outcome measures, clinical and biological information hindered analyses. Thus, while it is a commendable advancement that (some) trials now allow researchers to study their data, we conclude that gaining access to past trial datasets is complicated, frustrating the field's communal effort to find the best treatments for the right individuals. We provide a plea to promote harmonization and open access to data, by urging trial sponsors and the academic research community alike to remove barriers to data access and improve collaboration through practicing open science and harmonizing outcome measures, to allow investigators to learn all there is to learn from past failures and successes. HIGHLIGHTS: Pooling data from past Alzheimer's disease clinical trials may help identify subgroups most likely to benefit from specific treatments and may help inform future trial design.Accessing past trial datasets is complicated, frustrating the field's communal effort to find the best treatments for the right individuals.We urge trial sponsors and the academic research community to remove data access barriers and improve collaboration through practicing open science and harmonizing outcome measures.

Longitudinal cerebrospinal fluid measurements show glial hypo- and hyperactivation in predementia Alzheimer's disease.

BACKGROUND: Brain innate immune activation is associated with Alzheimer's disease (AD), but degrees of activation may vary between disease stages. Thus, brain innate immune activation must be assessed in longitudinal clinical studies that include biomarker negative healthy controls and cases with established AD pathology. Here, we employ longitudinally sampled cerebrospinal fluid (CSF) core AD, immune activation and glial biomarkers to investigate early (predementia stage) innate immune activation levels and biomarker profiles. METHODS: We included non-demented cases from a longitudinal observational cohort study, with CSF samples available at baseline (n = 535) and follow-up (n = 213), between 1 and 6 years from baseline (mean 2.8 years). We measured Aß42/40 ratio, p-tau181, and total-tau to determine Ab (A+), tau-tangle pathology (T+), and neurodegeneration (N+), respectively. We classified individuals into these groups: A-/T-/N-, A+/T-/N-, A+/T+ or N+, or A-/T+ or N+. Using linear and mixed linear regression, we compared levels of CSF sTREM2, YKL-40, clusterin, fractalkine, MCP-1, IL-6, IL-1, IL-18, and IFN-γ both cross-sectionally and longitudinally between groups. A post hoc analysis was also performed to assess biomarker differences between cognitively healthy and impaired individuals in the A+/T+ or N+ group. RESULTS: Cross-sectionally, CSF sTREM2, YKL-40, clusterin and fractalkine were higher only in groups with tau pathology, independent of amyloidosis (p < 0.001, A+/T+ or N+ and A-/T+ or N+, compared to A-/T-/N-). No significant group differences were observed for the cytokines CSF MCP-1, IL-6, IL-10, IL18 or IFN-γ. Longitudinally, CSF YKL-40, fractalkine and IFN-γ were all significantly lower in stable A+/T-/N- cases (all p < 0.05). CSF sTREM2, YKL-40, clusterin, fractalkine (p < 0.001) and MCP-1 (p < 0.05) were all higher in T or N+, with or without amyloidosis at baseline, but remained stable over time. High CSF sTREM2 was associated with preserved cognitive function within the A+/T+ or N+ group, relative to the cognitively impaired with the same A/T/N biomarker profile (p < 0.01). CONCLUSIONS: Immune hypoactivation and reduced neuron-microglia communication are observed in isolated amyloidosis while activation and increased fractalkine accompanies tau pathology in predementia AD. Glial hypo- and hyperactivation through the predementia AD continuum suggests altered glial interaction with Ab and tau pathology, and may necessitate differential treatments, depending on the stage and patient-specific activation patterns.

Discovery of novel CSF biomarkers to predict progression in dementia using machine learning.

Providing an accurate prognosis for individual dementia patients remains a challenge since they greatly differ in rates of cognitive decline. In this study, we used machine learning techniques with the aim to identify cerebrospinal fluid (CSF) biomarkers that predict the rate of cognitive decline within dementia patients. First, longitudinal mini-mental state examination scores (MMSE) of 210 dementia patients were used to create fast and slow progression groups. Second, we trained random forest classifiers on CSF proteomic profiles and obtained a well-performing prediction model for the progression group (ROC-AUC = 0.82). As a third step, Shapley values and Gini feature importance measures were used to interpret the model performance and identify top biomarker candidates for predicting the rate of cognitive decline. Finally, we explored the potential for each of the 20 top candidates in internal sensitivity analyses. TNFRSF4 and TGF [Formula: see text]-1 emerged as the top markers, being lower in fast-progressing patients compared to slow-progressing patients. Proteins of which a low concentration was associated with fast progression were enriched for cell signalling and immune response pathways. None of our top markers stood out as strong individual predictors of subsequent cognitive decline. This could be explained by small effect sizes per protein and biological heterogeneity among dementia patients. Taken together, this study presents a novel progression biomarker identification framework and protein leads for personalised prediction of cognitive decline in dementia.

Biomarker A+T-: is this Alzheimer's disease or not? A combined CSF and pathology study.

The biological definition of Alzheimer's disease using CSF biomarkers requires abnormal levels of both amyloid (A) and tau (T). However, biomarkers and corresponding cutoffs may not always reflect the presence or absence of pathology. Previous studies suggest that up to 32% of individuals with autopsy-confirmed Alzheimer's disease show normal CSF p-tau levels in vivo, but these studies are sparse and had small sample sizes. Therefore, in three independent autopsy cohorts, we studied whether or not CSF A+T- excluded Alzheimer's disease based on autopsy. We included 215 individuals, for whom ante-mortem CSF collection and autopsy had been performed, from three cohorts: (i) the Amsterdam Dementia Cohort (ADC) [n = 80, 37 (46%) Alzheimer's disease at autopsy, time between CSF collection and death 4.5 ± 2.9 years]; (ii) the Antwerp Dementia Cohort (DEM) [n = 92, 84 (91%) Alzheimer's disease at autopsy, time CSF collection to death 1.7 ± 2.3 years]; and (iii) the Alzheimer's Disease Neuroimaging Initiative (ADNI) [n = 43, 31 (72%) Alzheimer's disease at autopsy, time CSF collection to death 5.1 ± 2.5 years]. Biomarker profiles were based on dichotomized CSF Aß1-42 and p-tau levels. The accuracy of CSF AT profiles to detect autopsy-confirmed Alzheimer's disease was assessed. Lastly, we investigated whether the concordance of AT profiles with autopsy diagnosis improved when CSF was collected closer to death in 9 (10%) DEM and 30 (70%) ADNI individuals with repeated CSF measurements available. In total, 50-73% of A+T- individuals and 100% of A+T+ individuals had Alzheimer's disease at autopsy. Amyloid status showed the highest accuracy to detect autopsy-confirmed Alzheimer's disease (accuracy, sensitivity and specificity in the ADC: 88%, 92% and 84%; in the DEM: 87%, 94% and 12%; and in the ADNI cohort: 86%, 90% and 75%, respectively). The addition of CSF p-tau did not further improve these estimates. We observed no differences in demographics or degree of Alzheimer's disease neuropathology between A+T- and A+T+ individuals with autopsy-confirmed Alzheimer's disease. All individuals with repeated CSF measurements remained stable in Aß1-42 status during follow-up. None of the Alzheimer's disease individuals with a normal p-tau status changed to abnormal; however, four (44%) DEM individuals and two (7%) ADNI individuals changed from abnormal to normal p-tau status over time, and all had Alzheimer's disease at autopsy. In summary, we found that up to 73% of A+T- individuals had Alzheimer's disease at autopsy. This should be taken into account in both research and clinical settings.

CSF proteomic signature predicts progression to Alzheimer's disease dementia.

INTRODUCTION: Individuals in the Alzheimer's disease (AD) continuum with mild cognitive impairment (prodromal AD) are at increased risk to develop dementia. Still, underlying pathophysiological processes remain unclear. We studied whether cerebrospinal fluid (CSF) proteome changes are related to time to clinical progression in prodromal AD. METHODS: We measured 671 CSF proteins in 49 prodromal AD individuals (67±7 years old, 22 [45%] female) from the Amsterdam Dementia Cohort. Associations of protein levels with time to dementia onset were tested with Cox regression models, followed by biological pathway enrichment analysis. RESULTS: Eighteen (36%) individuals developed dementia during follow-up. In total, 128 (98%) proteins were associated with a 1.4- to 17-fold increased risk of progression to dementia (all P < .05). These proteins showed enrichment for immune system processes, signal transduction, neuronal death, and neurodevelopmental biology. DISCUSSION: CSF proteome changes related to rate of progression to dementia can be detected in prodromal AD, providing more insight into processes involved in early AD pathophysiology.