Does a weak grip mean I should train it?
5 kg lower grip went with higher death risk (HR 1.20 women, 1.16 men), but it is a marker in a cohort, not a target; it does not show that training grip extends life.
Main caveat: Baseline-only exposure: one grip reading; no change over time.
Grip was measured once at recruitment, 2007-2010, and linked to deaths, hospital admissions and cancer registries. Associations were linear and were reported per 5 kg lower grip. They held after models adding height, illness and lifestyle, and after dropping events in the first two years. Stronger associations appeared at younger ages.
In a post hoc comparison, grip beat systolic pressure and self-reported activity per SD. The authors state they are after prediction, not causal inference.
Protocol
Who: 502 628 recruited (5.5% response to about 9 million letters); 502 293 with grip data; Table 1 describes 477 074.
Test: handgrip dynamometer, mean of right and left, absolute kg.
Follow-up: mean 7.1 (range 5.3-9.9) years for mortality; 6.1 (4.4-9.0) for incidence. 13 322 deaths (2.7%).
Models: sociodemographics, height, prevalent illness, lifestyle (smoking, BMI category, self-reported activity, sedentary time, diet), then a landmark model excluding the first 2 years. People with disease at baseline were not excluded from all-cause models.
Funding and interests: UK Biobank funders (Wellcome Trust, Medical Research Council and others), independent of the analysis; none declared.
Key Results
| Per 5 kg lower grip, HR (95% CI), fully adjusted (results text: landmark model, first 2 years of events excluded) | Women | Men |
|---|---|---|
| All-cause mortality | 1.20 (1.17-1.23) | 1.16 (1.15-1.17) |
| Cardiovascular mortality | 1.19 (1.13-1.25) | 1.22 (1.18-1.26) |
| Respiratory mortality | 1.31 (1.22-1.40) | 1.24 (1.20-1.28) |
| COPD mortality | 1.24 (1.05-1.47) | 1.19 (1.09-1.30) |
| All-cancer mortality | 1.17 (1.13-1.21) | 1.10 (1.07-1.13) |
| Colorectal cancer mortality | 1.17 (1.04-1.32) | 1.18 (1.09-1.27) |
| Lung cancer mortality | 1.17 (1.07-1.27) | 1.08 (1.03-1.13) |
| Breast cancer mortality | 1.24 (1.10-1.39) | — |
| Prostate cancer mortality | — | 1.05 (0.96-1.15) |
Prediction, added to age, sex, diabetes, BMI, systolic pressure and smoking: C index 0.722 to 0.735 for all-cause mortality (change 0.013, 0.011-0.015); change 0.012 (0.007-0.017) for cardiovascular mortality and 0.009 (0.007-0.010) for incident cardiovascular disease. Validation half: 0.012, 0.011, 0.009.
Per 1 SD change, post hoc (grip read as lower): grip 1.48 all-cause and 1.57 cardiovascular mortality; systolic pressure 1.03 and 1.26; self-reported activity 1.06 and 1.07. Activity by questionnaire against grip by dynamometer is not a fair race (inference).
Incident cardiovascular disease, per 1 SD change: grip 1.29, systolic pressure 1.30.
Weakness cut-offs (26 kg men, 16 kg women): higher hazard for most outcomes.
Reporting. Men's cardiovascular mortality is 1.22 in the abstract and discussion but 1.24 in the results text. Cut-offs read < in the abstract, ≤ in the results and ≥ for the reference group. The abstract's list of exceptions applies prostate cancer to both sexes; the results say men. One C-index upper limit prints as 0.913 beside 0.800, probably a slip (our reading). One sentence compares the weakest quarter with the lowest group, presumably meaning the highest.
Mechanisms
Measured: grip only, plus bio-impedance body composition in description.
Authors' interpretation: grip is a valid marker of overall limb strength and is tied to nutrition, activity, socioeconomic status and genes (they cite a 52% heritability); the mechanisms are probably multifactorial and could not be determined.
Inference: illness lowers grip and raises death risk at once, so low grip can be a consequence of what kills. A 2-year landmark narrows this and cannot close it, as the authors say.
Practical Application
Does a weak grip mean I should train it? 5 kg lower grip went with higher death risk (HR 1.20 women, 1.16 men), but it is a marker in a cohort, not a target; it does not show that training grip extends life.
Good for: a cheap reading of general muscle status, and a reason for a general health check when grip is clearly low. The authors propose it for risk screening where blood tests are unavailable, and say further work is needed.
Not good for: programming. In a randomised trial at retirement age, handgrip fell alike across groups by year 4, three years after a supervised year of heavy lifting ended (REF-1139), so grip may be a poor scoreboard for such training (Inference).
Not good for: individual prediction, where the C-index gain is small, or ranking grip above blood pressure or activity, a post hoc comparison.
Limitations
Baseline-only exposure: one grip reading; no change over time.
Reverse causation and residual confounding: the authors say neither can be excluded and the associations may not imply causality.
Volunteer bias: response 5.5%; healthier and less deprived than the UK at large (authors), though they hold the sizes of associations generalisable.
Short follow-up for a mortality study, in people 40-69.
Self-reported activity, sedentary time and diet in the models; diet for a subset of 211 064.
No training, no athletes.
How this entry was checked and changed
Why this level
large observational cohort, read from the full text. Huge n and a measured exposure do not lift it above the observational ceiling, and the paper's own aim is prediction rather than cause