Measure: Know Where You Stand

Each 11 lb (5 kg) lower grip strength came with a 16% higher rate of death (HR 1.16, 95% CI 1.13 to 1.20) across 139,691 people in 17 countries. Measured in: 139,691 adults aged 35 to 70 across 17 countries, 58% women (the PURE study). What could explain it instead: Observational. Low grip strength is often an early sign of illness rather than its cause, which the short four-year follow-up makes more likely rather than less.. Grip is a marker of overall condition rather than a target in itself, so training your hands will not move the outcome. The often-repeated line that grip beats blood pressure as a predictor is narrower than it sounds: it was a post-hoc comparison, measured per standard deviation, and it holds for death but not for new cardiovascular disease, where blood pressure predicted better. Median follow-up was only four years, and these were adults from 35 upward rather than an older-adult sample.

Every 0.1 m/s faster usual walking speed came with about 12% lower mortality (HR 0.88, 95% CI 0.87 to 0.90). Measured in: 34,485 community-dwelling adults aged 65 and over, pooled from nine cohorts, about 60% women. What could explain it instead: Observational, and slow walking reflects existing disease as much as it forecasts new disease.. A pooled analysis of individual participant data from nine selected cohorts rather than a systematic review, so it carries none of the publication-bias checks that phrase usually implies. Samples were mostly white and US-based. Predicted ten-year survival at 75 ran from 19% to 87% in men and 35% to 91% in women across the speed range, a remarkable spread for a test that takes seconds.

Against a correctly sized cuff, a regular cuff read 19.5 mmHg higher systolic (95% CI 16.1 to 22.9) in people whose arm required an extra-large cuff, 4.8 mmHg higher (3.0 to 6.6) in those requiring a large cuff, and 3.6 mmHg lower (-5.6 to -1.7) in those requiring a small cuff. Measured in: 195 community-dwelling adults in Baltimore across a range of mid-arm circumferences, mean age 54, 34% male, 68% Black, 51% with hypertension. Readings were taken with an automated oscillometric device in a research setting, so the figures isolate cuff error with every other variable controlled. At home the cuff error stacks with arm position and rest time rather than replacing them. Single-site and heavily Black, and mid-arm circumference distribution decides who is affected.

After adjustment for BMI, men in the highest fifth of waist circumference had 2.05 times the risk of death of those in the lowest (95% CI 1.80 to 2.33) and women 1.78 times (1.56 to 2.04). For waist-to-hip ratio the figures were 1.68 and 1.51. BMI remained significantly associated with death in models that also contained waist. Measured in: 359,387 participants from nine European countries in the EPIC cohort, mean follow-up 9.7 years, 14,723 deaths; lowest mortality risk sat at a BMI of 25.3 in men and 24.3 in women. What could explain it instead: Reverse causation through illness-related weight loss. Undiagnosed cancer, COPD and heart failure shrink the waist years before they kill, which loads the leanest group with sick people and steepens the apparent gradient.. Waist and BMI are each adding information the other misses, so this is a case for measuring both rather than replacing one. The models adjust for education, smoking, alcohol, physical activity and height, and cannot adjust away the underlying problem that illness causes weight loss before it causes death.

Risk-adjusted all-cause mortality fell across every fitness band with no upper limit found. The lowest fitness group carried 5.04 times the mortality of elite performers (95% CI 4.10 to 6.20), and below-average against above-average carried 1.41 (1.34 to 1.49). Those figures are comparable to or larger than coronary artery disease (1.29), smoking (1.41) and diabetes (1.40) in the same model. Measured in: 122,007 adults referred for symptom-limited exercise treadmill testing at a US tertiary center between 1991 and 2014, mean age 53.4, 59.2% male, median follow-up 8.4 years, 13,637 deaths. What could explain it instead: Reverse causation. Undiagnosed illness lowers treadmill performance before it becomes a diagnosis, and the least fit group in a referral population is enriched for people who were already unwell on the day they were tested.. Everyone in this cohort was referred for a treadmill test, which means everyone had a clinical reason to be there, so it is not a general-population sample. Fitness was estimated from peak metabolic equivalents on a treadmill rather than measured by gas exchange, and none of it can be reproduced with anything you own.

One in five could not hold the 10-second stance. Among them 17.5% died over the follow-up, against 4.6% of those who could, an adjusted hazard ratio of 1.84 (95% CI 1.23 to 2.78). Measured in: 1,702 adults aged 51 to 75 attending a private exercise-medicine clinic in Brazil, 68% men. What could explain it instead: Observational. Neurological and musculoskeletal disease impairs balance and independently raises mortality, and a clinic population differs from the general public in ways adjustment cannot remove.. Much the smallest of our three self-measures, and not a general-population sample: these were self-selected attendees at one clinic. The confidence interval is wide. Treat it as a free signal worth acting on rather than a precise multiplier, and not as equal in weight to the grip and walking-speed findings.

Of six health and function measures tracked quarterly for a year, only improved gait speed predicted eight-year survival. Mortality was 31.6% in those who improved by at least 0.1 m/s, 41.2% in those who improved transiently, and 49.3% in those who never improved (adjusted hazard ratio 0.42, 95% CI 0.29 to 0.61). Measured in: 439 adults aged 65 and over recruited through a Medicare health maintenance organization and Veterans Affairs primary care programs in the United States. What could explain it instead: Observational. Recovery from an acute illness or a hospitalization raises gait speed and independently improves survival, so improvement partly marks who was temporarily unwell at baseline rather than who got fitter.. This shows that people whose walking speed improves live longer, not that making your walking speed improve makes you live longer. The authors end by calling for research into whether interventions that raise gait speed affect survival, which had not been done. The sample is small, single-country, and drawn from two specific health systems.

In the same cohort where each 11 lb (5 kg) less grip carried a 16% higher rate of death, grip strength showed no significant association with incident diabetes, with hospital admission for pneumonia or COPD, with injury due to a fall, or with fracture. Measured in: 139,691 adults aged 35 to 70 across 17 countries, 58% women, grip measured with a Jamar dynamometer, median follow-up 4.0 years. What could explain it instead: Observational. Fracture and fall injury depend heavily on bone density, medication and home environment, none of which grip strength measures, so a null here reflects what the test does not read rather than an absence of relationship in the body.. A null over a median four years of follow-up is a weaker statement than a null over fifteen, particularly for outcomes as uncommon as fracture in a cohort starting at 35. The finding is that grip is a general marker rather than a specific one, and it does not license the reverse claim that hand strength is unrelated to falls.

Each entered alone, office pressure (hazard ratio 1.13 per 10 mmHg systolic) and home pressure (1.23) both predicted cardiovascular events. Entered into the same model, only home pressure still predicted events (1.22, 95% CI 1.09 to 1.37) and office pressure did not (1.01, 0.92 to 1.12). Systolic home pressure was the sole predictor of total mortality. Measured in: 2,081 randomly selected Finnish adults aged 45 to 74, followed a mean of 6.8 years, with 162 cardiovascular events and 118 deaths. What could explain it instead: Observational. People willing and able to self-measure at home differ in health literacy, medication adherence and comorbidity from those who are not, and that difference tracks cardiovascular risk independently of the reading.. 162 events is a modest number for a model comparison, and the confidence intervals overlap more than a single point estimate suggests. Home readings in this study were taken under instruction with a supplied validated device, which is not the same as readings taken on whatever monitor a person owns. Ambulatory monitoring, not home monitoring, remains the reference standard for out-of-office pressure.

Each 10 beats per minute higher resting heart rate carried a relative risk of 1.09 (95% CI 1.07 to 1.12) for all-cause mortality and 1.08 (1.06 to 1.10) for cardiovascular mortality. Against the lowest category, 60 to 80 beats per minute carried 1.12 and above 80 beats per minute carried 1.45 (1.34 to 1.57). All-cause mortality risk rose linearly from 45 beats per minute upward. Measured in: 46 prospective cohort studies in general populations, 1,246,203 participants and 78,349 deaths for all-cause mortality; 848,320 participants and 25,800 deaths for cardiovascular mortality. What could explain it instead: Observational. Fitness, thyroid disease, anemia, fever, subclinical heart failure and beta blocker use all set resting heart rate and independently affect mortality, and adjustment for traditional cardiovascular risk factors does not touch most of them.. The authors report substantial heterogeneity between studies and detected publication bias, which is why this sits at moderate rather than strong despite the size. Resting heart rate is also measured inconsistently across cohorts, and the cardiovascular risk increase only reached significance at 90 beats per minute.

Against people under 70 beats per minute at both measurements (8.2 ischaemic heart disease deaths per 10,000 person-years), those under 70 at first measurement and above 85 ten years later had an adjusted hazard ratio of 1.9 (95% CI 1.0 to 3.6), and those going from 70 to 85 up to above 85 had 1.8 (1.2 to 2.8). The association was not linear (p = 0.003 for quadratic trend), and a decrease in resting heart rate showed no general mortality benefit. Measured in: 13,499 men and 15,826 women in Norway without known cardiovascular disease, measured twice about ten years apart in the Nord-Trondelag health study, mean 12 years of follow-up, 3,038 deaths of which 388 from ischaemic heart disease. What could explain it instead: Observational. A resting heart rate that climbs over a decade is a signal of developing illness, weight gain, deconditioning or new medication, all of which raise cardiac mortality on their own, so the rise marks the process rather than causing it.. 388 ischaemic heart disease deaths across many change categories makes the individual hazard ratios imprecise, and the confidence interval on the largest of them touches 1.0. The finding that a falling rate carried no general benefit is the part most relevant to anyone trying to train the number down, and it is a null within an observational study rather than a trial of lowering it.

Across the top four quintiles of the Sleep Regularity Index against the least regular quintile, all-cause mortality was 20% to 48% lower, cancer mortality 16% to 39% lower and cardiometabolic mortality 22% to 57% lower. Sleep regularity predicted all-cause mortality more strongly than sleep duration did, on both model comparisons run. Measured in: 60,977 UK Biobank participants, mean age 62.8, 55.0% female, with more than 10 million hours of accelerometer data and 1,859 deaths over a mean 6.3 years. What could explain it instead: Observational. Shift work, chronic pain, depression and terminal illness all fragment sleep timing and independently raise mortality, so irregularity partly marks who is unwell or working nights.. Six years of follow-up in a cohort with a median regularity index of 81 leaves little room to separate irregular sleep from the shift work, illness and social circumstances that produce it. The comparison with duration rests on nested model tests whose p values (0.14 to 0.20) show duration adding nothing, rather than showing regularity to be a cause. UK Biobank participants are healthier and less deprived than the UK population.

Scored 0 to 10 from how many hand or knee supports you need to lower yourself to the floor and stand back up, lower scores carried higher mortality. Against the top band of 8 to 10, multivariate-adjusted hazard ratios were 5.44 (95% CI 3.1 to 9.5) for scores of 0 to 3, 3.44 (2.0 to 5.9) for 3.5 to 5.5 and 1.84 (1.1 to 3.0) for 6 to 7.5. Each one-point increase carried a 21% improvement in survival. Measured in: 2,002 adults aged 51 to 80, 68% men, attending the CLINIMEX exercise-medicine clinic in Rio de Janeiro, median follow-up 6.3 years, 159 deaths (7.9%). What could explain it instead: Observational. The conditions that make it hard to get off the floor, arthritis, obesity, and neurological and cardiac disease among them, independently raise mortality, so the score reads existing illness as much as it forecasts new illness.. The same self-selected clinic population as the ten-second balance finding, so it is not a general-population sample and the confidence intervals are wide. The score combines strength, flexibility and body weight, so a low result does not point to any one cause, and the test asks you to lower yourself to the floor, which is not safe to attempt alone if you are frail or have fallen in the past year.

The pooled survival analysis was built on a stopwatch and a 13-foot (4-metre) course: from a standing start, walk at your usual pace, as if walking down the street, with no further encouragement. Mean speed in the 34,485 participants was 0.92 m/s (SD 0.27). Measured in: 34,485 community-dwelling adults aged 65 and over, mean age 73.5, 59.6% women, 79.8% white, pooled from nine cohorts collected between 1986 and 2000. What could explain it instead: Observational. Slow walking is a common final pathway for heart failure, arthritis, neuropathy and depression, so the speed reads existing disease as much as it forecasts new disease.. The contributing cohorts did not all use a 13-foot (4-metre) course. Walk distances ran from 8 feet to 20 feet (6 metres) and were converted to a 13-foot (4-metre) equivalent by formula, including one conversion derived from just 61 people walking both distances. Reproducing the protocol at home gets you closer to the published figures than any variation of it, and the pooled number still carries that conversion inside it.

Against the arm supported on a desk at heart height, resting the hand on the lap overestimated systolic pressure by 3.9 mmHg (95% CI 2.5 to 5.2) and diastolic by 4.0 mmHg (3.1 to 5.0). An unsupported arm at the side overestimated systolic by 6.5 mmHg (5.1 to 7.9) and diastolic by 4.4 mmHg (3.4 to 5.4). Measured in: 133 adults aged 18 to 80 recruited in Baltimore, mean age 57, 53% female; 36% had systolic pressure of 130 mmHg or above and 41% had a BMI of 30 or above. Single-site trial with triplicate automated readings under research conditions, so it isolates arm position with everything else held constant. It does not say how often each position is used in practice, which is the quantity that decides the population-level error.

General health checks had little or no effect on total mortality (risk ratio 1.00, 95% CI 0.97 to 1.03; 11 trials, 233,298 participants, 21,535 deaths, high certainty, I-squared 0%) or cancer mortality (1.01, 0.92 to 1.12, high certainty), and probably little or none on cardiovascular mortality (1.05, 0.94 to 1.16, moderate certainty). Ischaemic heart disease events were unchanged (0.98, 0.94 to 1.03, high certainty). Measured in: 17 randomized trials, 15 reporting outcome data, 251,891 participants unselected for disease or risk factors; geriatric trials were excluded by design. This is about packaged screening of more than one organ system in people with no particular reason to be tested. It says nothing about a specific test ordered for a specific reason, nothing about established single-disease screening programmes, and nothing about older adults, who were excluded. Several included trials are old enough that the treatments available on a positive result have since improved.

For the 30-second chair stand, test-retest intraclass correlation was 0.84 in men and 0.92 in women, and chair-stand count correlated with maximum weight-adjusted leg press at r = 0.78 in men and r = 0.71 in women. Performance fell significantly across each decade from the 60s to the 80s and was lower in low-active than high-active participants. Measured in: 76 community-dwelling volunteers over 60, mean age 70.5, for the reliability and validity work; the criterion-referenced independence standards were later derived from 2,140 moderate-functioning older adults aged 60 to 94. What could explain it instead: Volunteer selection. People who agree to perform two maximal leg-press tests are fitter and more mobile than the population the test is used to screen, which inflates both the reliability and the correlation with leg press.. The validation sample was 76 generally active volunteers, and the authors limit their conclusion to generally active community-dwelling older adults. The published cut-offs describe the capacity associated with staying physically independent, which is a different question from whether you will fall, and neither paper measured falls.

Pooling receiver operating characteristic data, waist circumference improved discrimination of adverse cardiometabolic outcomes by 3% over BMI and waist-to-height ratio by 4 to 5% over BMI. Within-study comparison of the area under the curve found waist-to-height ratio significantly better than waist circumference for diabetes, hypertension, cardiovascular disease and all outcomes combined, in both sexes. Measured in: 31 papers covering more than 300,000 adults across several ethnic groups. These are discrimination statistics for detecting risk factors that are already present, not prospective prediction of events, and a 4 to 5% improvement in area under the curve is a modest gain over a measure most people already have. The included studies used different waist landmarks and different outcome definitions.

Across 120 studies and 236 samples, the measurement protocol had no substantial influence on the association between waist circumference and all-cause or cardiovascular mortality, cardiovascular disease or diabetes. Significant associations were found in 65% of samples, and the common protocols (minimal waist, midpoint between rib and hip, umbilicus) were distributed similarly among the significant and non-significant ones. Measured in: 120 studies, 236 samples, analyzed by an expert panel across sample size, sex, age, race and ethnicity. This says the site does not change the risk association at population level. It does not say the sites give the same number: they do not, and switching landmarks between your own measurements will produce a change that is entirely artefact. The review also notes that no scientific rationale has been published for any of the protocols recommended by major health authorities.

Average measured sleep was 6.0 hours against 6.8 hours reported, a gap of about 0.8 hours. Reports rose by only 31 minutes for each additional hour of measured sleep, and the correlation between reported and measured duration was 0.45. Modelled, people sleeping 5 hours over-reported by 1.3 hours and those sleeping 7 hours over-reported by 0.3 hours. Measured in: 669 middle-aged adults at the Chicago site of the CARDIA study, 82% of those invited, measured with three days each of wrist actigraphy, a sleep log and questions about usual sleep duration in two waves. What could explain it instead: Health, sociodemographic and sleep characteristics all altered the size of the discrepancy in this sample, so the over-reporting is not a constant offset that can be subtracted, and groups differing in health differ in how wrong their estimate is.. Actigraphy is itself an estimate: it infers sleep from movement and tends to score quiet wakefulness as sleep, so the true gap between belief and sleep may differ from 0.8 hours in either direction. One city, one age band, and three nights per wave.

Epoch-by-epoch sensitivity for detecting sleep was high across all devices (0.93 or better) while specificity for detecting wake was low to medium (0.18 to 0.54). Sleep stage comparisons were mixed, and devices performed worse on the disrupted-sleep night. Most devices matched or beat research actigraphy on sleep and wake; two Garmin devices performed worse. Measured in: 34 healthy young adults, 22 women, mean age 28.1, tested on three consecutive nights in a sleep laboratory including one disrupted-sleep condition, against polysomnography. Healthy young adults sleeping in a laboratory is close to the easiest condition a sleep tracker will ever face, and the devices already failed at detecting wake. Anyone with insomnia, apnea or fragmented sleep is outside what was tested. Consumer firmware is also revised continuously, so a device tested in 2021 is not the device sold now.

Across 158 publications covering nine commercial device brands, Fitbit, Apple and Samsung devices counted steps accurately in controlled laboratory conditions. Heart rate accuracy varied by brand, with Apple Watch and Garmin most accurate and Fitbit tending to underestimate. No brand was accurate for energy expenditure. Measured in: 158 publications examining nine commercial wearable device brands, in laboratory and free-living conditions. Accuracy figures come mostly from controlled laboratory testing, which flatters step counting: slow walking, pushing a trolley and wheelchair use are the conditions where counters fail, and they are underrepresented. Devices are redesigned faster than they are validated, so a brand-level verdict ages quickly.

Thirty practitioners rating ten tongue slides reached the 80% agreement threshold on 17.3% of occasions in the first session and 19.1% in the second, and almost all of those were simple yes-or-no questions: with more complex response choices, 80% agreement was reached 5% of the time. In a later study using a formal classification scheme, inter-rater reliability across 17 clinicians was moderate for tongue coating (Gwet AC2 0.49 to 0.55) and fair for tongue body color and other body features (0.34). Measured in: 30 traditional Chinese medicine practitioners rating 10 tongue slides across two sessions; separately, 17 clinicians in Hong Kong and mainland China rating 24 representative smartphone tongue images. What could explain it instead: Terminology rather than perception. Practitioners trained in different lineages and languages use different operational definitions and different tongue regions, so disagreement about a word is scored as disagreement about a tongue.. Both studies conclude that the problem is the definitions rather than the observation: agreement rose when practitioners used an explicit operating classification scheme, and the same-rater agreement between looking at a person and looking at the photograph was good to very good. This measures agreement about a description, and it does not test whether tongue signs carry clinical information.

At the usual threshold of 13.5 seconds or more, pooled specificity was 0.74 (95% CI 0.52 to 0.88) and pooled sensitivity 0.31 (95% CI 0.13 to 0.57), so the test misses roughly two thirds of the people who go on to fall. The review's own logistic regression found the score was not a significant predictor of falls (OR 1.01, 95% CI 1.00 to 1.02, p = 0.05). Measured in: 25 studies reviewed in community-dwelling older adults, of which 10 could be pooled for the meta-analysis. This is about predicting falls in community-dwelling older adults specifically. The test remains a valid measure of mobility, correlating with the Berg Balance Scale and with activities-of-daily-living scores in its original validation, and the review's conclusion is that it should not be used in isolation to identify people at high risk of falling rather than that it measures nothing.

Wearable users reported higher rates of symptom monitoring and preoccupation (p = 0.03) and more treatment concerns (p = 0.02) than non-users. 20% of wearable users experienced anxiety and always contacted their doctor in response to an irregular rhythm notification. After matching, atrial-fibrillation-specific health care use was significantly greater in users (p = 0.04), including more ECGs, echocardiograms and ablations. Measured in: 172 patients with atrial fibrillation at a US academic center, mean age 72.6, 42% women, of whom 83 used a wearable, followed over 9 months of merged survey and electronic health record data. What could explain it instead: Selection by health anxiety. People already preoccupied with their rhythm are more likely to buy a monitor and more likely to call a doctor, which would produce this entire association without the device causing anything.. Retrospective and propensity-matched, so the direction of causation is open: more anxious patients may be more likely to buy a wearable in the first place. 172 people in a US academic health system, and the authors call for prospective randomized study before concluding anything about net effect. This is a cardiac population, and it says nothing directly about readiness scores in healthy users.

Twelve eligible studies were found: three evaluated both intra- and inter-rater reliability and nine only inter-rater. Acceptable agreement was achieved where the method carried concrete operational definitions. Poor agreement traced to unclear definitions and terminology in the classical descriptions and to imprecise descriptions persisting in standardized systems. Most studies did not assess intra-rater reliability at all. Measured in: 12 studies of manual pulse diagnosis at the radial artery by human testers, published in English. A narrative review with no pooling, over a literature the authors describe as consistently limited by small samples and by testers who often knew the participants beforehand. It reports how far practitioners agree with each other, which is a prerequisite for a diagnostic method rather than evidence about what the method detects.

At twelve months, systolic pressure was 3.5 mmHg lower than usual care with self-monitoring alone (95% CI -5.8 to -1.2) and 4.7 mmHg lower with telemonitoring added (-7.0 to -2.4). Telemonitoring did not beat self-monitoring alone (-1.2 mmHg, 95% CI -3.5 to 1.2). Adverse events were similar across all three groups. Measured in: 1,182 patients over 35 with blood pressure above 140/90 mmHg, across 142 UK general practices, who were willing to self-monitor; 1,003 (85%) were included in the primary analysis. The effect came from GPs titrating medication on the readings, not from the readings existing, so self-monitoring with nobody adjusting anything is a different intervention with no reason to expect this result. Participants were selected for willingness to self-monitor and the trial was unmasked. The monitor manufacturer was among the funders.

Progressive resistance training improved gait speed by 0.08 m/s (24 trials, 1,179 participants, 95% CI 0.04 to 0.12) and produced a moderate to large improvement in getting out of a chair (11 trials, 384 participants, SMD -0.94, 95% CI -1.49 to -0.38). Muscle strength improved substantially (73 trials, 3,059 participants, SMD 0.84). Physical ability improved slightly (33 trials, SMD 0.14). Measured in: 121 randomized trials, 6,700 participants, mostly older adults, with training typically two to three times a week at high intensity. These are improvements in the test scores, not in the outcomes the test scores predict, and no trial here measured survival. Adverse events were poorly recorded across the literature, though joint pain and muscle soreness were commonly reported where they were tracked, and serious events were rare and not attributed to the exercise.