Reading the research · Study-design deep dive
Reading the Dog Aging Project design: which dogs supplied the evidence?
Start by asking which dogs actually produced the numbers you’re reading, because in the Dog Aging Project the answer changes depending on which cohort a given figure belongs to.
By Theo Marsh, Senior writer · 8 min read · Published · Updated
Which dogs are in the Dog Aging Project numbers?
Start by asking which dogs actually produced the numbers you’re reading, because in the Dog Aging Project the answer changes depending on which cohort a given figure belongs to. The precision cohort paper reports 976 dogs fully enrolled as of July 31, 2024, and that count sits inside a much larger participant pool: 47,408 DAP Pack dogs in the same baseline comparison table, of which 46,432 were not precision cohort members.
Those denominators aren’t interchangeable. In the 2025 Precision report, blood chemistry, flow cytometry and fecal metagenomics counts describe Precision dogs alone. Other DAP biological-sample studies can draw on a broader population. A claim built on owner-reported health history or geospatial environmental measures, for example, can pull from the much larger Pack.
The two source papers describe different moments in the program’s life. The 2022 Nature design paper describes intended structure and planned collections, citing over 30,000 canine participants and Pack age distributions as of December 2020 (DAP data release 1.0). The 2025 GeroScience Precision paper reports collections through its stated cutoff.
A useful habit when reading either: write down the date attached to every count before you compare it to anything else.
| Reference point | Population described | Date of count |
|---|---|---|
| 2022 design paper | Over 30,000 canine participants | 2022 publication |
| 2022 design paper, Fig. 1c | Pack size-stratified age distribution | December 2020 |
| Precision cohort paper | 47,408 total DAP Pack | Not stated (baseline comparison table) |
| Precision cohort paper | 976 precision dogs | July 31, 2024 |
The precision paper states the comparison between precision and non-precision Pack dogs is offered for reference only, since the precision cohort wasn’t designed to be representative of the Pack. Mean age illustrates why: 3.9 years in Precision versus 6.8 years in the Pack, a gap the authors attribute to deliberate recruitment of puppies.
So before you interpret any result, identify whether the analysis population is the Pack, the precision cohort, or a subset of precision dogs with a particular assay result.
Trace the household recruitment route
Owners nominated their own dogs. The papers describe volunteer-owner recruitment: the project drew no probability sample from a national dog register. The 2022 design paper describes recruitment through mainstream media, social media and word of mouth, with owners nominating dogs through the project website and then creating password-protected personal portals.
That route matters for how you read any population-level statement. The precision paper is explicit that direct recruitment of volunteer owners was chosen partly to reach US dog owners regardless of veterinary care choices or access, an aim that differs from enrolling through specific veterinary settings.
Pack entry is broader than Precision eligibility. Per the precision paper, the only requirements for DAP Pack participation are known or estimated age and residence in the 50 US states, excluding territories. An owner completes a brief Nomination Survey, sets up the portal, and completes all sections of the Health and Life Experience Survey (HLES).
The 2022 design paper adds a household constraint you should carry into any clustering assumption: participation was limited to one dog per household at that time, with no cap on total Pack size, and enrollment spanning all 50 states.
Self-selection follows directly from this design. Owner demographics in the precision paper show the shape of it: 94% white in both the precision cohort and the total Pack, and 22% of precision owners reporting yearly income of $180,000 or more. The authors name this as a limitation, noting overrepresentation of highly educated, high socioeconomic status, white owner participants across all census divisions, and stating that this may limit broad applicability.
Nothing in either paper claims the enrolled population mirrors US dog ownership. Treat Pack-wide percentages as descriptions of a volunteer sample whose owners had internet access, time for a long survey, and willingness to keep updating it annually for the dog’s lifetime.
Separate entry into the Pack from sampled-cohort selection
Joining the Pack and being chosen for a sampled cohort are two different events with two different mechanisms, and conflating them will distort how you read selection bias. Pack membership follows from completing the HLES. The 2022 paper describes planned random sampling within predefined strata. The later Precision report describes the invitation algorithm actually used; read those accounts separately.
The 2022 design paper lays out the intermediate gate: after submitting a qualifying electronic veterinary medical record, Pack dogs become eligible for the more deeply studied cohorts, where owners receive biospecimen collection kits. Strata were designed in advance for size, sex, age at enrollment, breed background and geography, and eligible dogs were randomly sampled within those strata until targets were met.
The Precision paper describes a weekly Cohort Eligibility Algorithm with five invitation criteria: owner consent to contact the primary care veterinarian, an uploaded VEMR deemed adequate for research, expressed willingness (somewhat or very likely) to participate in all precision activities, and open slots in both the primary strata (age, weight, sex) and the secondary stratum (census division).
Two consequences follow for interpretation.
First, the VEMR requirement filters toward dogs with documented veterinary care. The 2022 paper acknowledges this could bias sampled cohorts toward regularly cared-for dogs, while noting that access to regular care also supports studying dogs across the aging trajectory.
Second, the 2025 report says all eligible dogs with an available cell in the target strata were invited. Eligibility, open strata and consent shaped the enrolled group, a process that differs from a simple random draw across the entire Pack. Treat the 2022 planned-sampling description and the later cohort as separate claims: the later cohort was not a random cross-section of the full 47,408 dogs.
The precision paper counts 18,282 dog-owner pairs who consented to veterinarian contact, submitted adequate VEMRs, and expressed interest. From that group, 1,931 were invited on or before December 31, 2023.
Read the Precision sample counts in context
Enrollment attrition between invitation and full membership is documented step by step, and each step shrinks the denominator. In the 2025 Precision cohort paper, 1,315 of 1,931 invited owners (68%) completed Informed Owner Consent. 1,259 completed DNA collection. 976 completed the year 1 biospecimen kit and were enrolled as of July 31, 2024, slightly under the 1,000-dog goal.
Enrollment ran from January 8, 2021 through July 31, 2024. The first year 1 kit was collected July 6, 2021; the last on July 29, 2024. A targeted outreach campaign from June to August 2023 re-engaged non-responsive invitees and accounted for 3.8% (n=36) of participants, which over-enrolled some strata and under-enrolled others.
Assay-level counts sit below the kit count. From 976 year 1 kits: 936 CBC results (96%), 967 chemistry profiles (99%), 921 urinalyses (94%), 829 flow cytometry results (85%), 935 metabolomic results (96%), 944 fecal microbiome results (97%), and 816 RRBS results (84%).
| Assay | Results | Share of 976 kits |
|---|---|---|
| CBC | 936 | 96% |
| Chemistry profile | 967 | 99% |
| Urinalysis | 921 | 94% |
| Flow cytometry | 829 | 85% |
| Metabolomics | 935 | 96% |
| Fecal microbiome | 944 | 97% |
| RRBS | 816 | 84% |
Follow-up years carry their own eligibility denominators. As of July 31, 2024, 864 dogs were eligible for year 2 and 704 (81%) had submitted; 506 were eligible for year 3 and 380 (75%) had submitted. Those percentages describe eligible dogs at a single snapshot date under rolling enrollment. They say nothing about lifetime retention.
Missingness has named causes in the paper: inadequate collection volume, shipping disruption, and transit time, which the DAP has previously reported as significantly associated with sample missingness. Grossly hemolyzed plasma, at 500 mg/dL or greater on the CDC reference palette, is excluded from metabolomic preparation.
Life-stage fill rates also vary. Puppies reached 51 of a 60 target (85%), while mature adults reached 144 of 142 (101%).
Match each claim to the dogs who supplied it
Map a finding back to the actual analysis sample, using the Precision paper’s assay counts as a starting point. The assay constrains the available data, while analysis eligibility determines the final denominator. The report lists 816 year 1 RRBS results, 944 fecal microbiome results and 829 flow cytometry results. Those are available-result counts, and the sample size for any specific analysis can run smaller. A clock or microbiome study, for instance, may apply further exclusions or require additional paired measurements.
The same logic applies across cohorts. The 2022 design paper assigns cheek-swab low-pass whole-genome sequencing to all sampled cohorts (Foundation, Precision, TRIAD), with a planned n of 10,000. Biospecimen kits and activity monitors were assigned to Precision and TRIAD. Cardiology examinations were assigned to TRIAD alone.
Environmental data is one place where the boundary runs differently than readers expect. The 2022 paper lists geospatial data under cohorts included: all. Secondary sources cover tract-level sociodemographic variables, four gases and two aerosols for air pollution, county-level temperature and precipitation, and neighborhood walkability, attached by address. The precision paper confirms precision biospecimen data are linked to concurrent environment metrics. Environmental covariates aren’t exclusive to precision dogs.
Owner-reported instruments run Pack-wide. HLES and the Canine Social and Learned Behavior Survey are listed for all cohorts. The End-of-Life Survey covers enrolled dogs who die. Morphometrics, mobility and cognitive tasks were assigned in 2022 to Foundation, Precision and TRIAD, with Pack noted as a future addition.
A practical test when you meet a headline number: ask whether its denominator is a survey-completion count or a laboratory-result count. Those differ by an order of magnitude here.
One more distinction is worth preserving: the 2022 paper describes planned collections and targets, while the 2025 precision cohort report documents what was actually collected and processed through July 31, 2024. Don’t cite a 2022 plan as a completed dataset.
Keep the unanswered comparison visible
The Precision paper identifies a gap behind its national enrollment targets. Suitable population-based information about US companion dog ownership was not readily available to the authors, so the team pegged geographic targets to human population data from the 2020 Census across the nine US Census Divisions.
That substitution reflects a design choice about how the researchers set enrollment targets, distinct from any empirical finding about dog ownership patterns. It means enrollment balance reflects where people live, and the paper’s own geographic result shows residual imbalance: East South Central enrolled 44 dogs against a target of 60, while West North Central, Mountain and Pacific came in slightly over target.
Reproductive status was left unstratified because intact dogs are uncommon in the US, particularly among older dogs. The precision paper notes the literature on intact versus sterilized distribution in the US dog population is scant, and says the precision distribution appears to follow trends seen in other studies.
The Precision cohort included 73 intact females (8%) and 101 intact males (10%).
Two limitations affect longitudinal reading specifically. The cohort enrolled over roughly three years, so dogs experience differing social and environmental conditions across each cohort year, and year 3 kit collection was projected to finish in July 2026.
The paper notes that longitudinal biospecimen analyses are delayed while dogs complete the required collection periods; that statement should not be read as a current data-release status. Collection happened at each dog’s own primary care clinic.
The paper credits that with broad geographic reach and names the tradeoff: the DAP team can’t directly observe or perform collection, and adherence to written and video instructions isn’t measurable.
When you cite any DAP number, carry its date and its cohort label with it, and state which of the two papers reported it.