REF-1082 · Tanaka et al., 2001 · Journal of the American College of Cardiology, · Paper · PubMed

Age-predicted maximal heart rate revisited

A meta-analysis of group mean HRmax values (351 studies, 492 groups, 18,712 subjects) gave HRmax = 208 − 0.7 × age (r = −0.90), and a laboratory study measuring HRmax in 514 healthy subjects gave a virtually identical 209 − 0.7 × age, with no difference between men and women or by habitual activity; it is a population regression line, and the abstract reports no individual prediction error

B+moderate evidenceMeasured on non-to trained athletesWritten before October 2026, being re-checked

The authors set out to derive one general equation that predicts healthy adults' maximal heart rate (HRmax), since the common 220 − age formula had never been tested for validity in a sample containing enough older adults. Pooling group mean HRmax values in a meta-analysis yielded 208 − 0.7 × age; a tightly controlled lab study measuring HRmax in 514 healthy subjects cross-validated it (209 − 0.7 × age). Content from the abstract only.

Participants

Meta-analysis: group mean HRmax values pooled across 351 studies, 492 groups, 18,712 subjects.
Laboratory study: HRmax measured in 514 healthy subjects. Age range, sex split and fitness levels are not given in the abstract beyond "healthy adults", men and women, and "wide variations in habitual physical activity levels".

Protocol

Step 1: meta-analytic collection of group mean HRmax values and regression on age.
Step 2: cross-validation of the new equation in a laboratory-based study with measured HRmax.

Key Results

Meta-analysis: HRmax was strongly related to age (r = −0.90), with the equation 208 − 0.7 × age.
Laboratory study: regression equation 209 − 0.7 × age — virtually identical to the meta-analysis.
The regression line was not different between men and women, and not influenced by wide variations in habitual physical activity levels.
Authors' conclusions: age alone predicts HRmax to a large extent, regardless of gender and habitual physical activity; the equation in current use (220 − age) underestimates HRmax in older adults, which would lead to underestimating both the real physical stress of exercise testing and the appropriate intensity for prescribed exercise.
No individual prediction error (standard error, limits of agreement) is reported in the abstract. The r = −0.90 is from the meta-analysis of group means, which smooths away the spread between individuals.

Practical Application

"Can I just use 208 − 0.7 × age for my max heart rate?" Tanaka 2001 derived 208 − 0.7 × age from group means of 18,712 healthy adults (351 studies) and confirmed it in 514 measured healthy subjects — it is the best-known population estimate, and it beats 220 − age in older adults on average, but it is a population line: the abstract gives no individual error. It tells you the average for your age, not your maximum.
Not from this paper; consistent with SYNTH-38: do not anchor zones on an age formula, this one included. In 5069 endurance athletes tested to exhaustion, every age formula carried an individual RMSE of 9.1-10.5 bpm, and Tanaka specifically read significantly low in the runners (REF-3494) — about a zone's width of error for one person.
Use it only as a provisional, explicitly labelled estimate when nothing measured exists; replace it with a measured maximum or, better, a threshold heart rate from the athlete's own hard effort (SYNTH-38, SYNTH-40).
The finding that 220 − age underestimates HRmax in older adults matters for masters athletes: a 220 − age zone table will tend to set their zones too low on average. Inference, not tested here: the fix is to measure, not to switch formulas — REF-3494 found the bias between formulas smaller than the individual error inside each.
Not from this paper: Olympiatoppen's intensity-scale guide cites this study for the spread in maximal heart rate between individuals of the same age and sex; the size of that spread is not in the abstract.

Limitations

The meta-analytic equation is fitted to group means, so its r = −0.90 overstates how well age predicts one person's HRmax; the individual error is not given in the abstract.
Healthy adults, not endurance athletes; mode of exercise testing and criteria for a maximal effort are not described in the abstract.
The 2001 cross-validation sample (n=514) is described only as healthy; ethnic, age and sex composition are not in the abstract.
Cross-sectional: it describes HRmax across ages in a population, not how a given athlete's HRmax changes with age.
How this entry was checked and changed
Why this level

cross-sectional measurement study with ≥100 (n=514 laboratory cross-validation); the meta-analytic part pools group means, which describes the population line, not individuals