Chair Stands and an 8-Foot Walk Tracked 7-Year Mortality Risk

There is a moment that tends to arrive quietly, usually in a kitchen. A woman of seventy-four pushes up out of a chair and notices that her hand went to the table on the way, and that it did not used to. She does not mention it to anyone. Then a week of headlines says that fitness tests predict how long older adults live, and the coverage arrives with no number attached to it, which leaves the moment in the kitchen exactly where it was.

The study underneath those headlines is a Taiwanese national cohort of 13,423 people, and it is narrower and more specific than the wire copy suggested. It does not say that exercise is good, which needed no cohort. It says that of seven standardized fitness measurements, some carried far more information about the next seven years than others — and that the one carrying the most is the one least likely to be measured anywhere: how long it takes to get out of a chair, walk eight feet, turn around, come back and sit down.

Between 11 January 2015 and 25 November 2016, community-dwelling adults aged 65 or older across Taiwan completed a standardized battery of fitness assessments at testing stations, administered by certified examiners following the Senior Fitness Test protocol under nationally standardized procedures. Their records were linked to Taiwan's National Health Insurance data, and deaths were ascertained through the insurance death registry, with follow-up running to 31 December 2022. The analysis was carried out between July 2025 and May 2026 and appeared in JAMA Network Open in August 2026.

Seven assessments covered four domains. Cardiorespiratory fitness was a two-minute step test. Muscular strength was measured twice, as a 30-second arm curl and a 30-second chair stand. Flexibility was the back scratch and the chair sit-and-reach. Balance and agility were the one-leg stance and the eight-foot up-and-go, in which a person rises from a chair, walks eight feet, turns, returns and sits. The authors also built a composite index by summing each participant's sex-specific percentile rank across all seven tests with equal weighting, then splitting that index into fifths as well.

Each test was divided into sex-specific quintiles, with the poorest performers as the reference group — a deliberate choice, because treating the fifths as categories avoids assuming that the distance between them is even. Multivariable Cox proportional hazards models estimated adjusted hazard ratios, controlling for sociodemographic factors, recorded comorbidities and, in the detail that matters most, self-reported physical activity. The cohort had a mean age of 72.9 years (SD 6.1) and was 62.5% female, 8,394 women in all. Over a median follow-up of 7.0 years (IQR 6.7-7.1), 1,631 participants died, 12.2% of the group.

The gradient was orderly. Across every assessment and across the composite index, age-adjusted cumulative mortality climbed step by step from the best-performing fifth to the poorest, with no reversal in the middle of the distribution.

Measured against the lowest-performing fifth, the highest-performing fifth had adjusted hazard ratios of 0.41 (95% CI, 0.33-0.51) on the eight-foot up-and-go and 0.50 (0.42-0.59) on the one-leg stance — both balance and agility measures. Lower-body strength followed at 0.55 (0.46-0.65) for the chair stand, then cardiorespiratory fitness at 0.58 (0.49-0.68) for the two-minute step test. Upper-body strength was weaker at 0.63 (0.53-0.75) for the arm curl, and flexibility weakest of all at 0.79 (0.67-0.93) for the chair sit-and-reach, an interval whose upper bound comes within a whisker of no association. The composite of all seven beat every individual test, at 0.39 (0.32-0.48).

Split by sex, the ordering shifted slightly. Among men, balance and agility held the lowest hazard ratios, followed by cardiorespiratory fitness. Among women, lower-body strength ran about level with balance and agility, with cardiorespiratory fitness behind them. Sensitivity analyses that treated each assessment as a continuous percentile score, or as its original raw value, produced the same inverse associations, so the finding does not rest on where the quintile lines happened to fall.

Set deeper in the paper is the result the coverage might have led with, and also the one most likely to be misread. For four of the assessments the relationship was not a straight line. Fitting restricted cubic splines to raw values, the authors found that mortality risk fell steeply across the lower end of performance and then levelled off: at approximately 15 repetitions on the 30-second chair stand and around minus 10 centimetres on the back scratch among men, and at 7 to 8 seconds on the eight-foot up-and-go, plus the same back-scratch value, among women.

A plateau is not a pass mark. Those figures describe where the curve stopped falling in this particular cohort — the point past which further repetitions bought little additional separation in seven-year mortality. The study set no thresholds, defined no normal range, and never tested whether a person who moved from one side of a plateau to the other changed their risk. Read as targets they would be an invention. Read as shape, they say something worth knowing, which is that the steepest part of the association sits at the bottom of the distribution, among the people already doing worst.

Clinical risk stratification in older adults leans heavily on counting diagnoses. A person with four chronic conditions is flagged; a person with none is not, even if it takes them eleven seconds to stand up and cross a room. The authors' argument is that functional capacity is a separate axis of information about physiological reserve, and that the equipment for reading it is a chair, a stopwatch and eight feet of floor.

The strongest support for that argument is the adjustment nobody quotes. These models already controlled for self-reported physical activity — for the answer people give when asked how much they move — and objective performance still separated the fifths. Asking and measuring are not the same thing, and in this cohort only one of them carried the signal. What the authors did not do is demonstrate that fitness testing outperforms the comorbidity-based scores clinics already run; they state plainly that no formal prognostic comparison, no C statistic and no reclassification analysis, was attempted.

The finding worth keeping here is not that fitness is good for people. It is the ordering. Balance and agility — the cheapest domain to measure and the one least likely to appear in any chart — outranked lower-body strength, which outranked upper-body strength, which outranked flexibility, consistently, with intervals that do not overlap at the ends. That ranking is a real contribution, and it points at a test taking under a minute and costing nothing. The caution is equally plain, and the authors put it first themselves: this is an observational cohort assembled from people well enough to walk into a testing station and attempt seven physical tasks, measured on one occasion, with no smoking data. A design like that will always flatter fitness, because the frailest are partly selected out of it and partly mid-decline at the moment of measurement. So the honest summary is that performance on these tests marks who is at risk, which is a genuinely useful thing for a clinician to know, and that nothing in these 13,423 records shows that practising the test moves the risk it marks. Those are two different claims, and a stopwatch cannot tell them apart.