Study note — JAMA 2020 Saint-Maurice (daily steps)
Shelf: Siim-cited
Full-text access: Yes — open on PMC (not paywalled for our pass).
Fetched 2026-09-18 from https://pmc.ncbi.nlm.nih.gov/articles/PMC7093766/ (also JAMA Network; DOI resolves).
Labels below: [Paper] = stated in the article · [Inference] = our take for Product / Walking.
Citation
Saint-Maurice PF, Troiano RP, Bassett DR Jr, Graubard BI, Carlson SA, Shiroma EJ, Fulton JE, Matthews CE.
Association of Daily Step Count and Step Intensity With Mortality Among US Adults.
JAMA. 2020;323(12):1151–1160.
doi:10.1001/jama.2020.1382 · PMID 32207799 · PMC PMC7093766
Methods [Paper]
- Design: Prospective observational analysis of NHANES 2003–2006 accelerometer cohort linked to mortality through Dec 31, 2015.
- Sample: 4,840 US adults (representative weighting); mean age 56.8 y; 54% women; 36% with obesity.
- Exposure: Hip-worn ActiGraph 7164; mean wear 5.7 days, 14.4 h/day. Mean steps/day 9,124. Some 2003–2004 missing steps imputed.
- Step intensity: extended bout cadence, peak 30-min cadence, peak 1-min cadence (steps/min).
- Outcomes: Primary all-cause mortality; secondary CVD and cancer mortality (1,165 deaths over mean 10.1 y follow-up; 406 CVD, 283 cancer).
- Analysis: Cox models; cubic splines and quartile/category contrasts; adjusted for age, sex, race/ethnicity, education, diet, smoking, BMI, self-reported health, mobility limitations, and multiple diagnoses.
Key results (numbers) [Paper]
Unadjusted incidence density — all-cause mortality by step category
| Steps/day | n | Deaths | Incidence density |
|---|---|---|---|
| <4,000 | 655 | 419 | 76.7/1000 person-years |
| 4,000–7,999 | 1,727 | 488 | 21.4/1000 |
| 8,000–11,999 | 1,539 | 176 | 6.9/1000 |
| ≥12,000 | 919 | 82 | 4.8/1000 |
Adjusted contrasts vs 4,000 steps/day (all-cause)
| Contrast | HR (95% CI) | Rough relative reduction [Inference from HR] |
|---|---|---|
| 8,000 vs 4,000 | 0.49 (0.44–0.55) | ~51% lower hazard |
| 12,000 vs 4,000 | 0.35 (0.28–0.45) | ~65% lower hazard |
Also [Paper]: 2,000 vs 4,000 → HR 1.51 (1.41–1.62) — higher mortality at very low steps.
Intensity
[Paper]: Higher peak cadence looked protective before adjusting for total steps; after adjusting for total steps/day, e.g. highest vs lowest peak-30 cadence quartile HR 0.90 (0.65–1.27), P trend .34 — not significant. Same pattern for CVD/cancer mortality after step adjustment.
CVD / cancer (vs 4,000 steps/day) [Paper]
- CVD: 8,000 vs 4,000 → HR 0.49 (0.40–0.60)
- Cancer: 8,000 vs 4,000 → HR 0.67 (0.54–0.82)
Limitations / catch [Paper]
- Observational — no causal inference.
- Unmeasured / residual confounding; higher steps may reflect better overall health; PAD and comorbidity severity not fully controlled.
- ActiGraph counts ambulatory activity that isn’t only “walking”; misses non-ambulatory exercise (swim, cycle).
- Cadence metrics are clock-minute accumulations, not pure “minutes of continuous stepping.”
- Death-certificate cause may misclassify.
- Excluded people (missing data) differed from analytic sample — generalizability caveat.
- Baseline ~7-day wear; does not model step changes over years.
- Step counts can differ ≥20% across devices (discussion).
What we take for Walking [Inference]
- Publishing a floor near ~8–8.6k and an optimal near ~12k is consistent with the paper’s dose points (big drop by 8k vs 4k; further lower hazard at 12k vs 4k).
- Emphasize total daily steps over cadence/intensity for mortality framing.
- Dual units on the board: steps + minutes via Tudor-Locke ≥100 steps/min heuristic (steps ÷ 100); see Tudor-Locke cadence. Siim did not supply that conversion in corpus.
- Do not claim the paper proved “walking causes longevity” — association only.
vs Siim’s use
| Siim (site / framing) | Paper |
|---|---|
| “12,000 vs 4,000 → 65% lower mortality” | [Paper] HR 0.35 for 12k vs 4k → [Inference] 1 − 0.35 = 65% lower hazard — matches his headline math |
| “No significant observed benefits beyond 12,000” | [Inference] His summary. The article’s highlighted HRs are 8k vs 4k and 12k vs 4k; we did not find a published sentence that formally tests “beyond 12k vs 12k.” Unadjusted rates still fall from the 8–11,999 band (6.9) to ≥12k (4.8/1000 py). Treat “plateau beyond 12k” as Siim framing, not a verified paper claim |
| Steps as longevity foundation | Aligns with [Paper] dose–response direction; he soft-pedals intensity — also aligns with [Paper] null intensity after adjustment |
Paywall note
Not paywalled for this pass via PMC full text. JAMA Network HTML also serves the article; PMC was our working copy.