Evidence summary: GrimAge Epigenetic Test Kit (Clock Foundation)

The Clock Foundation is a non-profit that describes itself as a UCLA spin-out [6][7]. Its at-home kit collects a small blood sample, which the vendor describes as a finger-prick sample [3][4]. The product page lists a GrimAge biological age score, epigenetic "surrogate markers", personalised recommendations, raw data on request and a consultation [3]. The foundation's page for clinical-trial researchers says it reports both the original 2019 GrimAge and the 2022 GrimAge2 on every result set [8]. The kit page does not say which version the consumer report shows, and we found no sample report. The foundation's homepage and its clinic and home-phlebotomist versions also mention a "GrimAge Systems Aging" report; we found no published description of that method and could not confirm whether the at-home kit includes it [4][16][17]. The site names DunedinPACE among the tests it offers, but DunedinPACE is not listed for this kit or in the shop [3][4][19]. The pages we reviewed do not name the lab that processes the samples [3][4][8]. The foundation's own terms state that "epigenetic age estimates are algorithmic and are not definitive measures of health" [14].

The lab analyses DNA methylation, a set of chemical marks on DNA, in the blood sample [3]. GrimAge does not measure blood proteins directly. Its developers used methylation patterns to estimate seven blood proteins and cumulative smoking (pack-years), combined these estimates with age and sex in a model trained to predict time to death, and converted the resulting risk score into years [21]. In their words: "Technically speaking, DNAm GrimAge is a mortality risk estimator." [21] GrimAge2 adds methylation-based estimates of two routine blood tests, C-reactive protein and HbA1c [22]. The "surrogate markers" listed for the report are these methylation-based estimates, not lab measurements of the proteins themselves [3][21]. When the developers built GrimAge, only 12 of 88 proteins could be estimated with even moderate accuracy, and the selected estimates correlated only moderately with measured levels in test data [21].

GrimAge was developed and validated in blood samples collected in large research studies; these papers do not report using finger-prick samples [21][22]. The vendor pages we reviewed do not say whether the kit collects a dried blood spot or liquid blood [3][4]. In one small study of 20 older adults in Southeast Asia, average GrimAge from finger-prick blood collected in small tubes was similar to venous blood from the same people, but the paper reports group averages, not how closely each person's two results agreed [33]. Another study found similar GrimAge2 values in dried spots made from venous blood and in white-cell samples from the same children [34]. In the sources we reviewed, we found no peer-reviewed comparison of finger-prick dried blood spots with venous blood for GrimAge, and no published validation of this kit's own sample collection and lab processing.

GrimAge Epigenetic Age Test is currently rated Promising on Compare [1]. The rating concerns mortality-related risk from the GrimAge result. It does not cover the surrogate markers as individual laboratory values. This page explains that rating. It does not change it.

• Development by its creators. GrimAge was built using data from the Framingham Heart Study and then tested by its developers in about 6,900 people from five US and Italian cohorts [21]. In those data, a GrimAge one year above the expected value was linked to about a 10% higher risk of death during follow-up, and GrimAge predicted death better than earlier clocks [21]. The same team tested GrimAge2 in 13,399 blood samples from nine cohorts, where its associations were slightly stronger than the original's [22]. In 173 adults under 40 and in saliva samples, the developers found that GrimAge2 still tracked related health measures, but its values were systematically offset; in the under-40 group, they differed from chronological age by a median of 11 years [22].
• Findings in other cohorts. Research groups with no GrimAge developer among the authors and no declared conflicts of interest found that the original GrimAge was associated with death in a Swedish twin cohort followed for about 16 years and in a nationally representative sample of older Americans [24][25]. In Generation Scotland (18,859 people), GrimAge2 had the strongest link with death over ten years among 14 clocks, but the authors found no single best clock across diseases [26]. That study's senior author advises the Clock Foundation and co-wrote the GrimAge2 paper [22][26]. An Irish study of 490 people, 35 of whom died during follow-up, found GrimAge to be the only clock that predicted death; GrimAge's developers were co-authors [23]. A 2026 peer-reviewed paper, carried out under a research agreement with another test company, TruDiagnostic, ranked the principal-component version of GrimAge narrowly first among the clocks it compared for predicting five-year survival in the same Generation Scotland cohort [27]. Its authors also include the Clock Foundation adviser mentioned above [27].
• Who is behind it. The University of California owns the patent rights to GrimAge; Ake Lu and Steve Horvath are the named inventors [21][30][35]. Horvath co-founded the Clock Foundation, is a paid consultant to it and chairs its scientific advisory board; the foundation licenses these patents [7][22][35]. Two other GrimAge2 authors, Riccardo Marioni and Alexandra Binder, are advisers to the foundation [7][22]. The foundation is a non-profit that sells these tests [6][7]. We did not review the foundation's financial filings or licence terms.
• This kit. In the sources we reviewed, we found no published study of this kit's own results, covering its sample collection, its lab and its report [3][8].

Our methodology's Strong tier asks for independent replication across populations and laboratories, open methods and demonstrated value beyond standard risk factors, and it rates a commercial test on the evidence for that product's own output [2]. For the original GrimAge, the link with mortality has been reported by research groups with no GrimAge developer among the authors [24][25]. For GrimAge2, the largest outside analysis we reviewed had a Clock Foundation adviser as its senior author [26]. The larger gaps concern this test and the method's added value. We found no published validation of this kit's own results. Where added value was tested, it was modest: in Generation Scotland, adding GrimAge2 to age, sex, smoking, alcohol, body mass index, education and deprivation raised a standard measure of prediction accuracy for ten-year mortality from 0.851 to 0.865 [26], and in the US study, the authors found that for most outcomes, information that is easier and cheaper to collect, such as demographics, socioeconomic status, mental health and health behaviours, predicted as well or better [25]. The method is published but patented, and the pages we reviewed do not disclose the kit's lab and processing steps [3][8][22]. Our methodology also lists GrimAge2 as an example of Promising evidence for mortality-related risk [2]. Our proposed reading is that these gaps explain why the test sits below Strong under the current criteria. This is our editorial interpretation, not the original rating rationale.

In the sources we reviewed, two randomised trials reported GrimAge results, both in later analyses of stored blood samples. In CALERIE, 220 adults without obesity were assigned to two years of calorie restriction or their usual diet [29]. A version of GrimAge rebuilt to reduce technical noise (the principal-component version [28]) did not change differently between the two groups [29]. In DO-HEALTH, 777 Swiss adults aged 70 or older took part for three years [30]. GrimAge2 was slightly lower on average in people given omega-3 than in people who were not, while vitamin D and a home exercise programme showed no effect on any clock, and the principal-component GrimAge did not change significantly [30]. The DO-HEALTH authors, who include Steve Horvath, state that the significance of these changes for long-term survival is unknown [30]. Neither trial tested buying this kit or using it to guide care [29][30]. The results are mixed across trials and GrimAge versions. This summary is not a recommendation to take supplements or change your diet.

When the same blood sample is run twice, the original GrimAge gives very similar results, more so than older clocks [28]. A 2026 reanalysis also found GrimAge among the most repeatable clocks on repeat runs [31]. But when small groups had blood drawn several times within hours or days, around events such as a meal or a stress test, both GrimAge versions were among the less stable clocks, far less stable than their repeat-run results suggest; the principal-component version did somewhat better [31]. Without control groups, the authors could not tell whether the events or simply the timing of the draws caused this [31]. The study was small, mostly involved adults aged 18–27, used older array versions and compared repeat runs within single labs only [31]. Two of its authors are co-inventors of a competing clock, Systems Age, and have received consulting fees from TruDiagnostic [31]. For research trials, the Clock Foundation advises collecting at least two samples before and two after an intervention [8]. The vendor states a test precision of one to two years [4]; we found no data supporting that figure in the sources we reviewed. None of the sources we reviewed sets a threshold for a meaningful change in one person.

GrimAge results show associations with health and survival across groups of people [21][24][25][26]. A GrimAge result in years is an age-like score derived from mortality risk. It is not a measure of how fast you are aging now, and its developers note that a GrimAge eight years above the expected value does not mean eight fewer years of life [21]. The developers expect these measures to complement existing clinical tests, not replace them [21][22]. Experts on aging biomarkers separate whether a measurement is technically reliable, whether it predicts outcomes, and whether using it improves health [32]. The reviewed studies do not show whether using this test to guide decisions improves health. The vendor's terms say results are for research and information and are not intended to diagnose disease [14][15]. We do not treat this test as a diagnostic tool.

For current price and availability, see Compare. If you spot an error, please tell us.