Biofluid biomarker data

This section provides an overview of the biofluid biomarker data, which constitutes a significant and comprehensive portion of the dataset presented in ADNI studies. While the sheer volume and variety of the biofluid biomarker data tables might initially appear complex, this documentation aims to provide a clear structure and guide for effective navigation. 

The biofluid biomarker data tables can be understood and categorized in several ways: 

  • By Source: Some tables contain analyses conducted by ADNI-funded research teams (‘internal’ sources), reflecting studies performed within the ADNI framework. Other tables present results from analyses conducted by external laboratories on ADNI biological samples obtained through specific requests (‘external’ sources). 
  • By Biofluid Type and Analytical Platform: Tables are also distinguished by the biological fluid used for analysis, such as Cerebrospinal Fluid (CSF), plasma, serum, or urine. Furthermore, they are categorized by the analytical technology employed, including methods like mass spectrometry or immunoassay. 
  • By Biomarker Focus: A primary way to organize the tables is by the specific biomarkers measured. This includes core Alzheimer’s Disease (AD) biomarkers, such as proteins from the amyloid and tau families, as well as a broad spectrum of other biological markers that may not be exclusively related to AD. Additionally, some tables contain data best described under ‘omics’ approaches (like proteomics or metabolomics), providing comprehensive molecular profiles. 

By considering these different organizational perspectives – source, fluid type, analytical method, and biomarker focus – this documentation will help you understand the structure and content of the biofluid data files and effectively identify the data most relevant to your research needs. 

Biofluid sample collection

A prerequisite to any biofluid data being available is, naturally, the in-clinic collection of the relevant biofluid. The types of biofluids collected from ADNI participants, and the frequency of that collection, has varied historically throughout the ADNI phases. A table laying out the historical collection schedule is available on the ADNI website, and is reproduced here for convenience:

ADNI1

In ADNI1, all participants across the Cognitively Unimpaired (CU), Mild Cognitive Impairment (MCI), and Dementia (DEM) cohorts had biofluids collected on the same schedule.

Biofluid Collection Frequency
Urine all visits
Plasma from Blood all visits
Serum from Blood all visits
CSF1 baseline and month 12
Blood for NCRAD baseline
Urine screening
Urine baseline
  1. CSF sample collection contingent on participant consent to lumbar puncture

ADNIGO

Note that rollover participants from ADNI1 do not undergo a ‘baseline’ visit, with their first visit under the ADNIGO protocol being coded as an ‘initial’ visit. This is largely a matter of semantics in this particular case.

Biofluid Collection Frequency
CSF screening
Plasma from Blood all visits
Serum from Blood all visits
APOE/GWAS baseline/initial
RNA baseline/initial
Cell Immortalization baseline/initial

ADNI2

During ADNI2, the study schedule became more variable compared to previous phases. Consequently, the data from the ADNI2 cohort follows three distinct collection schedules, each specifically applying to a different subset of the participants enrolled during this phase.

Newly Enrolled CN and MCI Participants

Biofluid Collection Frequency
CSF baseline and every two years
Buffy Coat all visits
Plasma from Blood all visits
Serum from Blood all visits
RNA baseline, then yearly
APOE/GWAS baseline
Cell Immortalization baseline

Newly Enrolled DEM participants

Note that this is largely the same as the schedule for newly enrolled participants outside of the DEM cohort, however per protocol these participants are not subjected to annual follow-ups beyond a two-year window.

Biofluid Collection Frequency
CSF baseline and Month 24
Buffy Coat all visits
Plasma from Blood all visits
Serum from Blood all visits
RNA baseline, Months 12 and 24
APOE/GWAS baseline
Cell Immortalization baseline

Rollover Participants

Biofluid Collection Frequency
CSF 1st ADNI 2 visit, then every two years
Buffy Coat 1st ADNI 2 visit, then yearly
Plasma from Blood 1st ADNI 2 visit, then yearly
Serum from Blood 1st ADNI 2 visit, then yearly
RNA 1st ADNI 2 visit, then yearly
APOE/GWAS 1st ADNI 2 visit if not obtained in ADNI GO

ADNI3

For the ADNI3 phase, the biofluid collection schedules are more streamlined. Differences in scheduling occur solely based on whether a participant is newly enrolled or a rollover from a previous ADNI phase.

Newly Enrolled Participants:

Biofluid Collection Frequency
CSF baseline and every two years
Buffy Coat all visits
Plasma from Blood all visits
Serum from Blood all visits
RNA all visits
PBMCs baseline
APOE/GWAS baseline
Cell Immortalization baseline

Rollover Participants

Biofluid Collection Frequency
CSF baseline and every two years
Buffy Coat all visits
Plasma from Blood all visits
Serum from Blood all visits
RNA all visits
PBMCs baseline

ADNI4

The ADNI4 phase introduces further complexity in biofluid collection schedules. Variations in scheduling depend on several factors: participant enrollment status (newly enrolled versus rollover), diagnostic group, and, for rollover participants, the timing of their last PET scan conducted under the ADNI3 protocol.

Additionally, ADNI4 includes a process for remote blood collection. Select participants residing at a distance may utilize a local phlebotomy site for scheduled blood draws.



Newly Enrolled CN Participants

Biofluid Collection Frequency
CSF - optional baseline, months 24, 48
Buffy Coat baseline
Plasma and serum from Blood baseline, months 24, 48
RNA baseline
PBMCs baseline
APOE/GWAS baseline

Newly Enrolled MCI Participants

Biofluid Collection Frequency
CSF - optional baseline, months 24, 48
Buffy Coat all visits
Plasma and serum from Blood all visits
RNA all visits
PBMCs baseline
APOE/GWAS baseline

Newly Enrolled DEM Participants

Biofluid Collection Frequency
CSF - optional baseline, month 24
Buffy Coat baseline, months 12 and 24
Plasma and serum from Blood baseline, months 12, 24
RNA baseline, months 12, 24
PBMCs baseline
APOE/GWAS baseline

Finding specific analytes

One common question from researchers who are interested in working with ADNI biofluid biomarker data is “Where do I find X?”, where X is some specific biomarker that they’re interested in studying. These sorts of questions always feel lacking. A much better question is, “where do I find X, and what was the context of the analysis?” 

We’ll return to that question of context in a later section, as it deals heavily with the issue of external vs. internal analyses. For now, let’s focus on the more straightforward question of finding specific analytes in the mass of biofluids analysis tables. 

If a given analyte has been measured somewhere in the data set, there are two places in which its presence is recorded. 

  1. The data dictionary hosted on the IDA. Finding analytes in this table is usually a simple matter of searching for relevant keywords in the list of field descriptions. 
  1. Within the documentation of a specific data set. This is particularly the case for analytes that may have been captured in an -omics data set, which are generally accompanied by a separate data dictionary. 

In the latter case, finding whether a biomarker has been recorded can be a touch tricky, as it requires first identifying analyses where the analyte of interest might be found - for instance, if the analyte in question is a protein then it may have been included in a broad proteomics assay - and then downloading the specific files and working with the documentation and data dictionaries present therein. 

Internal and external analyses

Some analyses are carried out by the ADNI Biomarker Core, which is currently based at the University of Pennsylvania. As such, any file with ‘UPENN’ in the title is almost certainly an internal data set. Most of the other data sets are from external investigators working with ADNI samples. 

Why is this distinction relevant? It’s a matter of sampling bias. Assays carried out locally at clinical sites, or in funded ADNI labs are carried out across all ADNI specimens, while analyses from external groups may only be interested in a select subset of the ADNI cohort and are likely to request samples for analysis in a non-random fashion. 

For example, corporate labs interested in validating an assay are likely to request samples from ADNI participants who have amyloid PET scans available, as this provides a basis for validation. Academic labs interested in studying the longitudinal trajectory disease progression may only request samples from participants who have been enrolled in the study for a long time, etc. 

The specific criteria for sample requests are generally included in the documentation for a given data set, and investigators should be sure to review these documents carefully to determine whether the data is suitable for answering their questions of interest. 

Breaking down the biospecimen results files

At the time of writing, there are 216 files available for investigators to download in the Biospecimen Results subsection of the IDA. These tables, and their methods documents, are named in a fairly consistent fashion as illustrated by the figure below:

Now that we know how to read the listings, let’s begin by singling out the tables that data users should pay particular attention to:

Key tables

Out of the tables available in this section, there are a few that are particularly noteworthy, and should generally serve as the starting points for researchers interested in working with this data.

In general, these tables are both

  • comprehensive – the measurements recorded are generally carried out on biosamples from a majority of the ADNI cohort, rather than on a specific nonrandom subsample.

  • Important – the biomarkers in these tables are broadly considered to be major biomarkers of AD pathology, or important biomarkers in a general clinical sense

APOE Genotyping

The main APOE genotyping results table is produced by the ADNI Genetics Core lab at Indiana University. It can be found on the IDA underneath the Biospecimens -> Biospecimen results heading.

This table consists of a standard set of structural identifiers (Phase, PTID, RID, and VISCODE), and a field indicating APOE genotype. Additional information was also included during the ADNI1 phase, but these fields are not populated for any of the subsequent phases.

APOE is polymorphic, with three common alleles: ε2, ε3, and ε4. In the table, participant genotype is recorded in the aptly named GENOTYPE field, where it is written using the notation n/m, with n ≤ m, to indicate the presence of one εn allele and one εm allele.

For example, a GENOTYPE value of 3/3 indicates that the participant carries two ε3 alleles, while a GENOTYPE value of 3/4 indicates that the participant carries one ε3 allele and one ε4 allele. APOE ε4 homozygotes are indicated as 4/4, etc.

UPENN - Plasma AB42, AB40, ptau217, NfL, GFAP

Users who are interested in working with major plasma AD biomarkers should start here:



This is an internal data set based on analyses carried out by the ADNI biomarker core lab at the University of Pennsylvania. It contains longitudinal measurements for the ADNI4 cohort, as well as a subset of the cohorts from previous phases.

Methods


UPENN - CSF Biomarkers Roche Elecsys

Much like the plasma analyses mentioned above, this is a comprehensive, longitudinal data set produced by the ADNI biomarker core, using the Roche Elecsys immunoassay platform.

Molecules measured in this table include Amyloid beta 42 (Aβ42), Amyloid beta 40 (Aβ40), total tau, and p-tau181.

Methods

Important Note

CSF local lab results

This table consists of measurements of several important molecules in CSF – including protein, glucose, and white/red blood cell counts.

It contains information for all participants who underwent LPs, and the assays are carried out on-site.

URMC safety lab results

This table contains data on a number of important non-AD blood and urine biomarkers for participants in ADNI2, ADNI3 and ADNI4.

This is, notably, one of the only sources of information on important biomarkers of chronic diseases such as diabetes and renal dysfunction.

Note that this table is stored in ‘long’ format. The majority of ADNI tables are ‘wide’, in the sense that each row corresponds to a single participant and a single timepoint, with columns corresponding to measurements taken on that participant at the specified timepoint.

This is not the case in this table. Each participant and time point has multiple rows, each of which records the result of a single test.

Amprion alpha-synuclein seeding assay

Alpha-synuclein is an important biomarker of Parkinson’s disease and Lewy body dementia, which are important copathology in the context of AD.

This contains the results of a CSF seed amplification assay, evaluating the presence of misfolded alpha-synuclein. It contains cross-sectional results for more ADNI participants, based on their last-available CSF samples at the time of the analysis.

Methods

-omics data

The line between –omic and non-omic measurements is difficult to draw clearly. In terms of the data produced from -omics studies, and how we’ll be classifying data as –omics data for the purposes of this document, we generally look towards the following characteristics:

  • Measures an array of molecules based on their structure (proteomics, lipidomics), or function (metabolomics)

  • Large numbers of analytes, sometimes numbering in the hundreds or thousands. Because of this, -omics data sets tend to include their own separate data dictionary, with field descriptions not available in the general ADNI data dictionary.

Note that ADNI –omics that falls under umbrellas like genomics and transcriptomics is the purview of the genetics core and is not located in the Biospecimen Results section of the data repository.

The bulk of the –omics files available for download - in fact, the bulk of all biospecimen results files available for download - are a result of work carried out by the Alzheimer’s Disease Metabolomics Consortium (ADMC), an international consortium of researchers focused on understanding the role of metabolic changes in AD, who have made extensive use of ADNI biosamples in their investigations.

A summary of this project is provided as a methods document on the IDA, and is replicated here:

In addition to the ADMC data, a number of other labs have made –omics data available via LONI. A full listing of the –omics data sets is provided in the following table, along with some basic information about the data, and links to methods docs where they are available.

Table N PTs Longitudinal N records* N analytes Phases
CruchagaLab SomaScan7k proteomics 735 No 737 7008 1, GO, 2, 3
Biomarkers Consortium CSF Proteomics MRM Data 287 No 322 320 GO, 2
FNIH Biomarkers Consortium CSF Proteomics Project 198 Yes 750 278 1, GO, 2
Hu Lab CSF Inflammatory Proteins 386 No 398 14 1, GO, 2
Emory University CSF Targeted MS 833 No NA2 69 GO, 2
Biomarker Consortium CSF QC Multiplex3 311 No 327 159 ADNI1
Biomarker Consortium Plasma QC Multiplex 566 Yes 1063 190 ADNI1

*note that some cross-sectional may contain replicates, etc

1. Analysis also included samples from non-ADNI studies.

2. Data is stored in long format

3. VISCODE field is titled ‘visit_code’; RID field is listed as ‘rid’

4. VISCODE field is titled ‘Visit_code’

Other biofluid biomarker tables

Having covered the key tables of AD biomarkers and the various –omics tables, the remaining data consists of targeted measurements of AD and non-AD biomarkers in blood or CSF. The analyses represented here run the gamut from validation studies of experimental assays for well-known biomarkers, alternative methodologies for measuring biomarkers, assays targeting novel biomarkers, and the results of historical analyses.

Many of these tables represent results of validation studies conducted by the FNIH biomarker consortium. These studies were largely conducted with the purpose of validating novel assays using a select subset of ADNI samples. A more comprehensive source for data on the biomarkers listed in these tables is generally available in the primary UPENN data sets.

All other tables are summarized in the table below, with methods docs linked where available.

Table Biofluid Biomarkers N PTs Longitudinal N records* Phases
APOE glycosylation Plasma, CSF ApoE isoforms and glycoforms 188 No 366

1

2
Bateman Lab Plasma Abeta42/Abeta40 Ratio (2019/2022)1 Plasma Aβ42, Aβ40

199 (19)

359 (22)

 No (19) 
Yes (22)

457 (19)

742 (22)

1

GO

2
Araclon Biotech S.L. AB test40 and 42 Plasma Analysis Plasma

Free Aβ42,

Free Aβ40,

Total Aβ42,

Free Aβ40

 305 Yes 784

1

GO

2
Zhang Lab CSF Hemoglobin ELISA and PS129 CSF Phosphorylated alpha-synuclein 372 Yes 805

1

GO

2
Blennow Lab Longitudinal Plasma NFL Plasma NFL 1191 Yes 3762 1
EUROIMMUN CSF Beta-Amyloid CSF Aβ42, Aβ40 121 No 280

GO

2
Haass lab CSF sTREM2/PGRN2 CSF sTREM2, PGRN 1036 Yes 1878

1

GO

2
Emory DDE Analysis Summary Plasma 4’,4 DDE3 208 No 218 1
Blennow Lab CSF GAP-43 CSF GAP-43 804 Yes 1268

GO

2
Blennow Lab CSF NFL (ADNI1)4 CSF NFL 399 No 415 1
Blennow Lab CSF NG CSF NG5 399 No 415 1
Blennow Lab Plasma Tau Plasma Total tau 581 No 581 1
C2N Diagnostics PrecivityAD2 Plasma Aβ42, Aβ40, p-tau217, np-tau217, ApoE proteotype 232 No 232 4
Diadem - AlzoSure Predict Plasma U-p53AZ 563 Yes6 584 2
Fagan Lab - CSF Visinin-like Protein-1 CSF VLIP-1, YKL-40, SNAP-25, NRGN7 152 Yes 587

1

GO

2
Fujirebio CSF beta-amyloid ratio CSF Aβ42, Aβ40 423 No 442

GO

2
Fujirebio Lumipulse Plasma amyloid beta Feasibility Study8 Plasma Aβ42, Aβ40 121 No 514 2
Homocysteine Plasma Homocysteine 819 Yes 3390 1
Selkoe Lab Plasma oAB Plasma oligomeric Aβ 809 No 829
University of Gothenburg Longitudinal Plasma p-tau181 Plasma p-tau181 1191 Yes 3758 1 
GO 
2
UW Plasma EV Apogee Plasma L1CAM, Aβ42, Aβ40 300 No 300 1 
GO 
2
Oeckl and Otto Lab Serum Beta Synuclein Serum Beta Synuclein 476 Yes 900 1 
GO
Janssen Plasma p217+tau Simoa Assay9 Plasma p217+tau 121 No 130

2

3
Saladax Biomedical Summaries CSF Aβ42, total tau 263 Yes 393

1

GO

2
Selkoe Lab NT1 Tau Plasma NT1 Tau 810 No 840

2

3
UW - CSF Complement 3 & Factor H CSF C3, FH 390 No 390 1
Redox Reactive Autoantibodies Serum See  doc 90 No 90 1

*note that some cross-sectional may contain replicates, etc

1. Two studies conducted; one in 2019, one in 2022; consult methods docs for more details on the differences between the resulting data sets.
2. Contains results from two different assay platforms
3. Insecticide metabolite 4,4’- dichlorodiphenyldichloroethylene
4. Not to be confused with the Blennow lab longitudinal CSF NFL data from later phases
5. Neurogranin
6. Longitudinal data for 3 participants
7. Visinin-like protein-1, chitinase-3 like-1, synaptosomal-associated protein 25, Neurogranin
8. Not to be confused with the plasma data from the ADNI biomarker core lab at UPENN, which is also measured using the luminpulse essay
9. Used the same set of aliquots used in the FNIH sponsored Abeta42/Abeta40 method comparison study