Heart rate variability declines with age. That much the studies agree on. Umetani et al. (1998), using 24-hour recordings from 260 healthy people aged 10 to 99, found RMSSD fell to 47 percent of its second-decade value by the sixth decade and then stabilised. The tables below come from a larger study with a shorter window. Use them as orientation, not as a target.
“Furthermore, there were significant modifications according to age in nearly all of the domains.”
Voss and colleagues recorded five minutes of ECG in 1,906 healthy adults from the German KORA S4 cohort, lying on their backs after five to ten minutes of rest. The values are mean ± one standard deviation. They found no significant difference between women and men in RMSSD or SDNN in any decade.
RMSSD (Oura, Whoop, most overnight)
| Age | Women (ms) | Men (ms) |
|---|---|---|
| 25–34 | 42.9 ± 22.8 | 39.7 ± 19.9 |
| 35–44 | 35.4 ± 18.5 | 32.0 ± 16.5 |
| 45–54 | 26.3 ± 13.6 | 23.0 ± 10.9 |
| 55–64 | 21.4 ± 11.9 | 19.9 ± 11.1 |
| 65–74 | 19.1 ± 11.8 | 19.1 ± 10.7 |
Source: Voss et al. (2015), Tables 5 and 7. Five-minute supine ECG, daytime, in a study setting. Not overnight wearable averages.
SDNN (Apple Watch, ECG strips)
| Age | Women (ms) | Men (ms) |
|---|---|---|
| 25–34 | 48.7 ± 19.0 | 50.0 ± 20.9 |
| 35–44 | 45.4 ± 20.5 | 44.6 ± 16.8 |
| 45–54 | 36.9 ± 13.8 | 36.8 ± 14.6 |
| 55–64 | 30.6 ± 12.4 | 32.8 ± 14.7 |
| 65–74 | 27.8 ± 11.8 | 29.6 ± 13.2 |
In this dataset SDNN sits only modestly above RMSSD: about 1.1 to 1.3 times higher in the youngest decade, about 1.5 times in the oldest. That ratio belongs to the five-minute window, not to the metrics. Shaffer and Ginsberg (2017)note that recording length strongly influences time-domain values, and that 24-hour and short-term measurements “cannot substitute for each other.” Don’t convert your Apple Watch SDNN into an Oura RMSSD, or either of them into this table.
Why these tables mislead more than they help
The spread inside each band is wide. For women aged 25–34, one standard deviation either side of the mean RMSSD runs from about 20 to 66 ms. That is a threefold range, and it covers only the middle of the group.
“A number of methodological discrepancies underlined disparate values.”
Nunan’s review pooled 44 studies and 21,438 healthy adults, and found that much of the disagreement between published norms came down to how the data were recorded and cleaned. Your device adds its own method on top. A ring averages across the night; a watch samples short windows; the table above is five minutes lying down. Time of day moves the number too: Natarajan et al. (2020), analysing data from over 8 million Fitbit users, reported strong diurnal variation in HRV.
Two specific failure modes to watch for:
- The 30-year-old with an RMSSD of 25who panics because it’s below the table’s mean. In the 25–34 group, 25 ms sits within one standard deviation of the mean for both women and men. A population average is not a diagnosis.
- The 55-year-old with an RMSSD of 40 who feels validated because the number clears the table by more than one and a half standard deviations. It still says nothing about whether their own trend is heading down, and the trend is the part worth watching.
The actually useful comparison
Your seven-day rolling average versus your seven-day rolling average a month ago. Researchers who monitor athletes day to day make the same move: Plews et al. (2013) list appropriate averaging among the fixes for reading HRV as a daily monitoring tool. Seven days is our default, not a number from their paper.
We don’t give percentage thresholds. We found no study that validates a cut-off for a consumer wearable’s rolling average, and a threshold without a source is a guess dressed as a rule. Watch the direction and how long it lasts. A sustained fall you can’t explain is worth attention. It is not a verdict.