Phenotypic Age Calculator: PhenoAge from a Standard Blood Panel

Phenotypic age calculator: nine PhenoAge blood markers, the units people mix up, and how to read the result.

Written by the Nuvo Longevity editorial team at Firefly Labs LLC.

NUVO wordmark on a dark navy field with a cyan arc, for the phenotypic age calculator guide

A phenotypic age calculator estimates biological age from a standard blood panel. Levine phenotypic age, also called PhenoAge, uses your chronological age plus nine biomarkers: albumin, creatinine, glucose, C-reactive protein, lymphocyte percent, mean cell volume, red cell distribution width, alkaline phosphatase, and white blood cell count. To run a PhenoAge calculator you need those nine results in the units below, and your age in years.

Published 22 October 2026. Written by the Nuvo Longevity editorial team at Firefly Labs LLC, the developer of Nuvo. Coefficients below are taken from Levine et al., Aging (Albany NY) 2018, Table 1, and from the 25 February 2019 correction to Liu et al., PLOS Medicine. This page is educational and is not medical advice. A phenotypic age is a research number. It does not diagnose disease or replace a clinician.

People comparing epigenetic clocks, blood models, and wearable scores can start with what biological age scoring models measure. That page is the map of model types. This page is the hands-on PhenoAge how-to: the markers, the unit conversions, the published equation, and how to read one result.

What phenotypic age is

Phenotypic age is an age in years built from clinical chemistry, not from DNA methylation and not from a watch. Morgan Levine and colleagues developed it in NHANES III, a US sample with long mortality follow-up. They started from 42 clinical markers plus chronological age, used a penalized Cox model with 10-fold cross-validation, and kept nine biomarkers plus age. Those ten inputs went into a Gompertz mortality model. The 10-year mortality score from that model was then converted into years. The training sample with complete biomarkers was 9,926 adults. The paper is Levine ME, Lu AT, Quach A, et al. An epigenetic biomarker of aging for lifespan and healthspan. Aging (Albany NY). 2018;10(4):573–591 (PMID 29676998).

The same 2018 paper then trains a second measure, DNAm PhenoAge, by regressing that clinical age on blood DNA methylation. DNAm PhenoAge uses 513 CpG sites. A biological age blood test that reads methylation is that second measure, or another clock. The calculator on this page is the first measure: the clinical phenotypic age you can compute when the nine markers are already on a lab report.

Liu and colleagues applied the clinical measure in an independent sample, NHANES IV (1999–2010): 11,432 adults aged 20–84 years, plus 185 adults whose age was top-coded at 85, with 1,012 deaths ascertained through 31 December 2011. In that cohort, phenotypic age remained associated with mortality after adjustment for chronological age, including across age bands, disease counts, and health behaviors. See Liu Z, Kuo PL, Horvath S, Crimmins E, Ferrucci L, Levine M. A new aging measure captures morbidity and mortality risk across diverse subpopulations from NHANES IV. PLoS Med. 2018;15(12):e1002718.

The nine biomarkers

Levine’s Table 1 names the marker, the unit the weight expects, and a body-system label. The “what it reflects” column here is that physiology in plain language. Direction follows the sign of the published weight: a positive weight raises phenotypic age when the marker rises.

Biomarker Unit in the equation Often printed on a US report What it reflects
Albumin g/L g/dL A blood protein made in the liver. Lower albumin raises phenotypic age in this equation. Levine’s table labels it liver.
Creatinine umol/L mg/dL A breakdown product used to judge kidney filtration. Higher creatinine raises phenotypic age here. Levine’s table labels it kidney.
Glucose mmol/L mg/dL Fasting blood sugar. Higher glucose raises phenotypic age here. Levine’s table labels it metabolic. For the separate three-month average, see HbA1c and biological age.
C-reactive protein natural log of mg/dL mg/L on many high-sensitivity reports, sometimes mg/dL A protein that rises with inflammation. Higher CRP raises phenotypic age here. The equation logs it. Levine’s table labels it inflammation.
Lymphocyte percent % % The share of white cells that are lymphocytes. A higher percent lowers phenotypic age here. Levine’s table labels it immune.
Mean cell volume fL fL Average red blood cell size. Higher mean cell volume raises phenotypic age here. Levine’s table groups it with immune markers.
Red cell distribution width % % How much red blood cell size varies. Higher RDW raises phenotypic age here, and the weight is one of the larger ones. Levine’s table groups it with immune markers.
Alkaline phosphatase U/L U/L An enzyme tied to liver and bone. Higher alkaline phosphatase raises phenotypic age here. Levine’s table labels it liver.
White blood cell count 1000 cells/uL ×10³/µL or ×10⁹/L, which is the same number The white blood cell count. A higher count raises phenotypic age here. Levine’s table labels it immune.

Chronological age in years is the tenth input. It is not a blood biomarker, and it stays in the equation even though the result is also an age. That is how the model keeps a person near their birthday when the blood panel looks ordinary, and lets the panel pull the result away from the birthday when it does not.

Units people convert wrong

The weights were fit on the units in the table. A correct marker in the wrong unit is a different number, and the age moves with it. These are the conversions the equation needs.

  • Albumin. g/L = g/dL × 10. A report of 4.3 g/dL is 43 g/L. Enter 43.
  • Creatinine. umol/L = mg/dL × 88.4. That factor is creatinine’s molecular weight (113.12 g/mol): 1 mg/dL = 88.4 umol/L. A report of 0.90 mg/dL is about 80 umol/L.
  • Glucose. mmol/L = mg/dL ÷ 18.018. That factor is glucose’s molecular weight (180.16 g/mol). A fasting glucose of 90 mg/dL is about 5.0 mmol/L. Enter about 5, not 90.
  • C-reactive protein. The weight multiplies the natural log of CRP in mg/dL. mg/dL = mg/L ÷ 10. A high-sensitivity result of 1.5 mg/L is 0.15 mg/dL, and the equation uses ln(0.15), which is negative. Logging 1.5 instead of 0.15 skips that divide-by-10. A CRP printed as 0 cannot be logged. A result below the assay’s limit is a different fact from typing the limit in as if it were a measured value. The papers do not publish one universal floor for that case.
  • White blood cell count. “1000 cells/uL” means the number already scaled by 1,000. A report of 6.2 ×10³/µL, or 6.2 ×10⁹/L, enters as 6.2. Multiplying by 1,000 again turns a normal count into nonsense.
  • Lymphocyte percent, mean cell volume, RDW, alkaline phosphatase. Percent stays a percent (32, not 0.32). Mean cell volume stays in fL. RDW stays in percent. Alkaline phosphatase stays in U/L when the report uses U/L.

Glucose in this equation is the chemistry value from the draw. Liu and colleagues required at least 8 hours of fasting before they calculated phenotypic age in NHANES IV, because glucose was one of the inputs. HbA1c is a different marker, an average over roughly three months, and it is not one of the nine.

The published formula

Use the equation from the correction, not the incomplete equation in the original PLOS Medicine methods. On 25 February 2019 the authors published the missing step: Liu Z, Kuo PL, Horvath S, Crimmins E, Ferrucci L, Levine M. Correction. PLoS Med. 2019;16(2):e1002760. The linear weights match Levine’s Table 1, with one rounding difference called out under the table. “ln” is the natural logarithm.

First sum a linear predictor, xb:

Term Coefficient Multiply by
Intercept −19.907 1
Albumin −0.0336 g/L
Creatinine 0.0095 umol/L
Glucose 0.1953 mmol/L
ln(CRP) 0.0954 natural log of CRP in mg/dL
Lymphocyte percent −0.0120 %
Mean cell volume 0.0268 fL
Red cell distribution width 0.3306 %
Alkaline phosphatase 0.00188 U/L
White blood cell count 0.0554 1000 cells/uL
Chronological age 0.0804 years

Levine’s Table 1 prints the alkaline phosphatase weight as 0.0019. The 2019 correction prints 0.00188. The worked example below uses 0.00188. On that made-up panel the two weights change the result from 46.95 years to 46.96 years.

Second, turn xb into the 10-year mortality score inside the model. The correction writes it as:

M = 1 − exp( (−1.51714 × exp(xb)) / 0.0076927 )

The constant 0.0076927 is the Gompertz rate (γ) from the same correction, and 1.51714 is the compact form of (exp(120 × γ) − 1), the 120-month horizon. M is an intermediate score inside the equation. It is a step in the arithmetic, and it is not a personal prognosis.

Third, convert that score into years:

Phenotypic age = 141.50 + ln( −0.00553 × ln(1 − M) ) / 0.09165

CRP has to be above zero so ln(CRP) exists, and M has to stay below 1 so the outer logarithm exists. A glucose value entered in mg/dL can push xb so high that M reaches 1 and the age step has no real answer. That failure is a unit error, not a biological age.

Illustrative worked example

The panel below is invented so the arithmetic can be checked. It is not a patient, not a target range, and not a result from any app. Every input is already in the formula’s units.

Input Value used About what a US report might show
Albumin 43 g/L 4.3 g/dL
Creatinine 80 umol/L 0.905 mg/dL (80 ÷ 88.4)
Glucose 5.0 mmol/L 90 mg/dL (5.0 × 18.018)
CRP 0.15 mg/dL 1.5 mg/L
Lymphocyte percent 32 32%
Mean cell volume 90 fL 90 fL
Red cell distribution width 13.2% 13.2%
Alkaline phosphatase 68 U/L 68 U/L
White blood cell count 6.2 6.2 ×10³/µL
Chronological age 52 years 52 years

ln(0.15) = −1.89712. Summing the correction’s terms gives xb = −8.752245. Then M = 0.030700. Phenotypic age = 46.95 years. On this made-up panel the blood age sits about five years under the chronological age of 52.

The same invented inputs, with one unit left unconverted, show why the conversions matter:

  • Albumin entered as 4.3 instead of 43 moves the result from 46.95 years to 61.14 years.
  • Creatinine entered as 0.90 instead of 80 moves it to 38.75 years.
  • CRP entered as 1.5 (the mg/L figure) instead of 0.15 moves it to 49.35 years.
  • Glucose entered as 90 instead of 5.0 pushes M to 1, and the age step is undefined.

Those four numbers are the equation applied to bad units. They are still illustrative. They are there so a real report is converted before it is typed in.

How to interpret the result

The output is an age in years. Set it next to chronological age on the day of the draw. A phenotypic age under the birthday means this panel looks younger than the age the model expects. A phenotypic age over the birthday means the panel looks older. In the papers, the research version of “acceleration” is PhenoAgeAccel: the residual from a linear regression of phenotypic age on chronological age in a population. A single person’s phenotypic age minus their birthday is the comparison you can see on one report. It is related to that residual, and it is not the same statistic as a regression fit on NHANES.

Read the markers that moved xb, not only the final year. In this equation, higher albumin and a higher lymphocyte percent pull the age down. Higher creatinine, glucose, CRP, mean cell volume, RDW, alkaline phosphatase, white blood cell count, and chronological age pull it up. RDW’s weight (0.3306 per percentage point) is large next to the others, so a one-point RDW difference moves the linear predictor more than a one-unit change in alkaline phosphatase.

Liu and colleagues also tied a higher phenotypic age to a higher count of diseases in NHANES IV, after accounting for chronological age. That is an association in a survey cohort. One calculator result still does not name a disease, rank a treatment, or tell one person how long they will live.

Limitations

Phenotypic age is one fasting blood draw, converted with one published equation.

  • It is a snapshot. The nine markers can move with hydration, a recent meal if you were not fasting, a new medication, or lab-to-lab method differences. The next draw is a new snapshot.
  • Acute inflammation skews CRP, and often the white cell count. A cold, an injury, a dental infection, or hard training in the days before the draw can lift CRP. The log then lifts xb, and the phenotypic age rises for a reason that may be gone in two weeks. A baseline question belongs to a draw taken when you are well.
  • The model was trained on NHANES. NHANES III (1988–1994) supplied the fit. NHANES IV supplied the validation Liu reported. Both are US civilian samples, with the age ranges and exclusion rules in those papers, including the fasting rule for glucose. Using the same coefficients outside that population is an extrapolation. The NHANES IV paper tests subgroups inside that survey. It does not certify the number for every lab, every country, or every assay.
  • The marker list is short on purpose. HbA1c, lipids, blood pressure, VO2 max, heart-rate variability, and sleep are outside the nine. A panel ordered for another reason may still be missing lymphocyte percent, RDW, or mean cell volume even when glucose and creatinine are present. Missing one input means you do not have this equation’s result.
  • A CRP of zero breaks the log. So does an out-of-range unit that saturates M. Both are reasons to stop and check the report, not reasons to invent a substitute constant.

How PhenoAge compares with epigenetic clocks and wearable scores

DNAm PhenoAge, from the same Levine 2018 paper, predicts the clinical phenotypic age from 513 methylation sites. It needs a methylation assay. It is the clock version of this blood age, and the number it returns is an estimate of the clinical age, not a re-run of the nine-marker arithmetic on that day’s chemistry.

Other methylation clocks answer different questions. Horvath’s 2013 multi-tissue clock uses 353 CpGs and was built to track chronological age across tissues (Genome Biology, 2013;14:R115). DNAm GrimAge (Lu AT, Quach A, Wilson JG, et al., Aging, 2019;11(2):303–327) combines methylation surrogates of seven plasma proteins with a methylation surrogate of smoking pack-years, and was built around time-to-death. A kit that reports one of those clocks is a methylation test, not this calculator. Price, sample type, and which clock is on the report differ by company. The shared limit is cadence: a methylation result arrives when you send a sample, then it sits still until the next kit.

Wearable scores move on a different schedule. Resting heart rate, heart-rate variability, VO2 max, sleep, and activity update on days when you never visit a lab. They describe function and recovery. They do not contain albumin, RDW, or alkaline phosphatase. A wearable biological age and a phenotypic age on the same morning are two measurements. The model-type comparison, including what each one can and cannot see, stays on the scoring models page so this one can stay on the arithmetic.

Consumer “biological age blood test” offers often bundle a panel with a proprietary model. PhenoAge is one published model you can compute yourself when the nine markers are on the report and the units match. What belongs in a longevity panel, and how often people repeat one, is covered in longevity blood test panels and rate of aging. Ways to test from home, including what a mail-in kit actually returns, are in how to test biological age at home.

How to track phenotypic age over time

Treat each result as a row in a log, not as a verdict.

  1. Use the same nine markers and the same conversions every time. Write the unit the lab printed next to the unit you entered.
  2. Draw blood in a similar state: fasting if glucose is on the panel, and when you are not in the middle of an infection or the days after a significant injury.
  3. Record the draw date, chronological age that day, xb if you want to audit the math, phenotypic age, and the simple gap versus chronological age.
  4. Compare rows with each other. A single gap of a few years can be the draw. A gap that keeps the same sign across two or three well-timed panels is the pattern worth talking about with a clinician.
  5. Expect blood cadence to be slow. Many people repeat a chemistry panel about once or twice a year, or on the schedule their clinician sets. Daily movement between draws comes from other inputs, such as sleep and fitness, which this equation does not read.

Where Nuvo fits

Nuvo Longevity (App Store: Nuvo: Longevity Coach, id 6762535065) is an iPhone and Apple Watch app. It reads HealthKit, including Apple Watch and, when those devices write into Apple Health, Oura or Garmin. Optional blood labs can be added. From those inputs it calculates a daily biological age from 21 biomarkers, on the phone. Coaching runs as on-device AI. There is no account. Apple’s privacy nutrition label for Nuvo is Data Not Collected. The price is $4.99/month or $24.99/year after a 7-day free trial. $24.99/year is $2.08/month, billed annually.

Nuvo’s daily score is the app’s own 21-biomarker calculation. The nine-marker equation on this page is the published PhenoAge calculator for a blood panel you compute yourself. Optional labs you already have, such as fasting glucose, CRP, or creatinine, can sit in Nuvo’s daily score alongside sleep, fitness, and recovery. The phone recalculates that daily number between blood draws. The phenotypic age you compute from the correction above is a separate result, updated when you have a new panel.

Download app. For a comparison of biological age apps, see the 2026 app roundup. For the difference between epigenetic, blood, and wearable models, stay with scoring models.

Sources

Frequently asked questions

What is a phenotypic age calculator?
A phenotypic age calculator turns chronological age and nine routine blood biomarkers into an age in years, using the Levine phenotypic age equation, also called PhenoAge. The inputs are albumin, creatinine, glucose, C-reactive protein, lymphocyte percent, mean cell volume, red cell distribution width, alkaline phosphatase, white blood cell count, and age.

Which units does the PhenoAge calculator expect?
Albumin in g/L, creatinine in umol/L, glucose in mmol/L, the natural log of C-reactive protein in mg/dL, lymphocyte percent, mean cell volume in fL, red cell distribution width in percent, alkaline phosphatase in U/L, white blood cell count in 1000 cells/uL, and age in years. Many US reports print albumin in g/dL, creatinine in mg/dL, glucose in mg/dL, and CRP in mg/L. Those values have to be converted before they enter the equation.

Is phenotypic age the same number as an epigenetic clock?
Clinical phenotypic age is the blood-chemistry calculation on this page. DNAm PhenoAge is a DNA methylation clock trained to predict that clinical age from 513 CpG sites, described in Levine et al., Aging, 2018. Horvath’s 2013 clock and GrimAge are other methylation measures. A methylation kit and this nine-marker calculator are different tests.

Does Nuvo run the Levine phenotypic age equation?
Nuvo calculates a daily biological age from 21 biomarkers on the iPhone, using HealthKit data from Apple Watch, Oura, or Garmin, plus optional blood labs. That daily score is Nuvo’s own on-device calculation. The nine-marker equation on this page is the published PhenoAge calculator for a blood panel.

How should I track a biological age blood test over time?
Repeat the same nine markers with the same unit conversions, in a similar fasting state, and outside an acute illness that would raise CRP. Record the phenotypic age next to chronological age on that date. The pattern across repeats is the signal. One draw is a snapshot.

Is a phenotypic age result a diagnosis?
A phenotypic age result is a research measure. Levine and Liu reported associations with mortality and morbidity in NHANES cohorts. It does not diagnose a disease, set a treatment target, or tell one person how long they will live. This page is not medical advice.