NOVOS Age vs. Other Biological Age Clocks
We cover every dimension that matters when comparing biological age clocks.
| Attribute | NOVOS Age Clock | Saliva-based Clock by Celebrity Scientist's New Startup | Other tests |
|---|---|---|---|
| Tissue Collection | Blood from a small poke of a finger, a method that is more accurate than via saliva | Saliva from a cheek swab, a method that is generally not very accurate | Blood collection methods that are invasive and far more uncomfortable that via small pokes of fingers |
| Sample Size | Samples from more than 20,000 humans | Samples from more than 8,000 humans | Samples typically from fewer than 2,500 humans |
| DNA Methylation Technology | Built using the modern MethylationEPIC array that measures 850,000 DNA sites and tests your sample on that same technology | Built using the modern MethylationEPIC array that measures 850,000 DNA sites but does not test your sample on that same technology | Built using older arrays that only capture 27,000-450,000 DNA sites |
| Chronological Age Range | 8-102 years | 18-100 years | Less expansive age range often lacking individuals 90+ years of age |
| Diversity | Significant diversity across ethnicity, race, and sex, all supported by many peer reviewed publications | Diversity across ethnicity, race, and sex, but without support of peer reviewed publications | Insufficient representation across ethnicity, race, and sex and without support of peer reviewed publications |
| Test Reliability | Optimized to be reliable across repeat measurements, with published and peer reviewed best-in-class ICC values (accuracy) >.96 for all three algorithms | Claims of being optimized to be reliable across repeat measurements without disclosing ICC values | Exhibit high test-retest error rates |
| Model Type | 3rd generation (latest) clock, the only one trained on longitudinal analysis (people across their lifetimes), the best way to track biological age as shown in publications and tested via peer review | Self-claimed "novel" method-based model that lacks publications, peer review, and head-to-head comparisons against other clocks | 1st generation (oldest) model trained to simply estimate chronological age instead of biological aging |
| Outputs and Analysis | Three: 1) 3rd generation Pace of Aging via DunedinPACE, 2) Biological Age, and 3) Telomere Length | One: A less accurate output of biological age | A single, less accurate output of biological age |
| Creators of Clock | A top team of Duke University and Columbia University scientists with peer reviewed publications | A start up company without publication of the algorithms, thus lacking scientific scrutiny | |
| Number of Studies | 45+ published studies by 30+ longevity scientists' labs across the world | Zero published studies | |
| Immune Cell Controls | Published and Patented Advanced 12-cell immune deconvolution methods (cell changes won't impact accuracy, which is common in saliva and makes blood samples better) | No controls | No controls |
| Studies that prove accuracy in different ethnic groups | Algorithms validated in the Family and Community Health Study of African American Families study, MESA (Multi-ethnic Study of Atherosclerosis), Cebu Longitudinal Health and Nutrition Survey (CLHNS Phillipines), Northern Finland Birth Cohort 1966 Study, Health and Retirement Study, the Normative Aging Study, the Framington Heart cohort, TILDA (the Irish Longitudinal Study of Aging), and many more. | No studies | No studies |
| Studies that show relatonship to outcomes | Algorithms have been validated in the Health and Retirement Study, the Normative Aging Study, the Framington Heart cohort and more. | No studies | No studies |
| Studies that show change with validated anti-aging interventions | The only algorithm proven to respond in a significant way to validated anti-aging interventions such as caloric restriction (Published in Nature) | No studies | No studies |
| Include Clinical Covariates | 21 clinical covariates and telomere length | No clinical covariates | No clinical covariates |
| Comparisons to other algorithms | Comparisons in the FHS study and in the Health and Retirement Study show superior results | No published comparisons | No published comparisons |
| Shares actual data on precision (ICC values) | See ICC values with comparisons in the FHS study. | No data | No data |
