In a network meta-analysis of 61 randomized trials with 4,744 participants, whole soy ranked best for lowering fasting glucose (SUCRA 91.0%), fasting insulin (95.4%) and HOMA-IR (83.4%), with isolated isoflavones next, while no soy product changed HbA1c. Rankings show relative order only and do not give the size of benefit.
In 12 randomized trials (n = 688), behavioural sleep interventions lowered systolic blood pressure by 4.91 mmHg versus control, and sleep extension by 7.59 mmHg. Heterogeneity was high (I2 = 74%) and heart rate variability did not change.
Across 16 studies, wearable-measured HRV was higher early in the menstrual cycle and lower toward the end, with time-domain differences of 3 to 9%, and was lower in hormonal contraceptive users. Studies defined cycle phases and menopause differently, so the results could not be pooled statistically.
In 318,000 US adults, sleeping 5 hours or less was associated with a 14.1 percentage point higher depression rate than sleeping 6-8 hours (95% CI 13.8%-14.4%), and sleeping 9 hours or more with a 12.9 point increase. These are self-reported, survey-based associations, so they do not show that sleep length caused the differences.
Among 2,680 undergraduates in Nanchang, China, negative life events were linked to more problematic smartphone use, partly through lower self-efficacy (indirect effect β = 0.05, 95% CI 0.04-0.07), and higher physical activity weakened the link (interaction β = -0.01, p < 0.001). This is a one-time self-report survey, so it cannot show that exercise or self-efficacy cause less phone overuse.
Across 27 randomized trials with 2022 children and adolescents, exercise lowered anxiety symptoms (SMD -0.36, 95% CI -0.58 to -0.15, p = 0.0009), with a larger effect in trials that met American College of Sports Medicine (ACSM) recommendations (SMD -0.67) than in the rest (SMD -0.13, not significant). The adherence grouping was applied by the authors afterwards across varied trials, so the dose comparison is only suggestive.
In a randomized trial of 62 people with hypertension, 8 weeks of AI-assisted remote rehabilitation improved 6-minute walk distance by 62.77 more than conventional offline exercise guidance (95% CI 26.33-99.22; P=.002) and lowered systolic blood pressure by 4.11 more (P=.046). The trial was small, open-label and short, and the systolic result barely reached significance.
In 59,078 UK Biobank adults, the best tertiles of sleep, activity and diet together were linked to 9.35 more years of lifespan than the worst, and a small combined step (5 min/day more sleep, 1.9 min/day more MVPA, 5 more diet points) to 1 more year. These are modelled life-table estimates from an observational cohort with 8.1 years of follow-up, so they are not proven gains from changing habits.
A scoping review of 17 studies of exercise for cognitive frailty found a wide mix of exercise formats, nearly half delivered digitally, and seven studies suggesting benefits through brain structure, inflammation, oxidative stress, neurotransmitters and blood flow. A scoping review maps what exists rather than measuring effect size, so it cannot say how much exercise helps.
A narrative review of peer-reviewed studies concluded that physical activity improves children's executive functions, attention, memory and academic performance, most consistently with cognitively engaging and moderate-to-vigorous activities. Effects varied with exercise design and intensity, and the abstract gives no study count or effect sizes, so the size of the benefit is unknown.
This narrative review found moderate-intensity exercise and the Mediterranean and DASH diets were linked to more diverse, favourable gut microbiota, while ketogenic diets brought fast weight loss but often fewer short-chain fatty acid-producing bacteria. It mixes human and animal studies of varied quality and gives no pooled numbers, so the conclusions are qualitative.
In 30 studies of wearable sensors in clinical and real-world settings, 67% (20/30) used inertial measurement units, and 4 studies (13%) reported high accuracy for fall risk and mobility decline (AUC up to 0.97). The studies were mostly observational, with varied sensor placement and methods, so the best-case accuracy may not hold in everyday use.
Across 26 studies, wearable approaches combining several signals had a pooled accuracy of 81.94% for detecting anxiety, versus 76.85% for single signals, with electrocardiography the best single signal (80.34% across 12 studies). Few single-sensor studies and varied methods mean the comparison is rough, and accuracy in research settings may not carry over to daily life.
Among 71,542 French women (845 with Parkinson's disease), high adherence to the Mediterranean or MIND diet was not linked to Parkinson's overall, but was linked to lower incidence under age 71 (HR about 0.75 for both). The confidence intervals reach 1.00 for one diet, and the finding comes from a subgroup, so it is suggestive.
In 3,752 Chinese university students, grit was positively associated with physical activity (β = 0.370, p < 0.001), and self-efficacy accounted for 21% of that effect. The survey is cross-sectional and self-reported, so it cannot show that grit or confidence leads to more activity.
In 967 older adults from three Chinese cities followed for one year, exercise frequency predicted a later sense of autonomous values (β = 0.166), and links between activity and well-being grew denser (56.84% to 60.00%). The data are two survey waves analysed as networks, so the paths show statistical prediction, not proof that exercise changed well-being.
In a survey of 391 college students, more physical activity was associated with lower social anxiety (β = -0.187, p < .001), and the sequential path through family support then self-efficacy accounted for 13.35% of the total effect. The survey was a single snapshot, so it cannot show that exercise caused the lower anxiety or that the paths run in this order.
In a meta-analysis of 15 studies with 698 healthy adults, SDNN, RMSSD, pNN50, HF and LF were all lower after acute high-altitude exposure than at sea level (all p < 0.001), while the LF/HF ratio rose (p < 0.001). The studies were small and varied, and the abstract gives no pooled effect sizes.