Counts describe this library, not the strength of evidence.
Cómo interpretar la evidencia
Sleep health has several dimensions. Regularity cohorts and sleep-extension trials answer different questions; a prediction from observational data is not proof that changing a score prevents disease. Exercise effects also differ between sleep quality and daytime sleepiness.
Across 22 articles (8 in the meta-analysis), Fitbit models without sleep staging overestimated total sleep time (effect size = -0.51, P<.001) and underestimated wake after sleep onset (effect size = 0.60, P<.001) compared with polysomnography (PSG). Sleep-staging models did better at detecting sleep, but their specificity (0.58-0.69) was still limited.
Among 1152 pregnant women in China, poor sleep quality in the second trimester (PSQI of 5 or more) was associated with antenatal depression (OR 3.42, 95% CI 2.48-4.72) and postpartum depression (OR 2.40, 95% CI 1.58-3.64). Sleep and mood were measured by questionnaire, so the link may run both ways and does not show that improving sleep prevents depression.
Among 11,875 US children aged 9 to 10, more time watching TV or movies was associated with 5.9% more rule-breaking behavior (IRR = 1.059), and each longer-sleep step with 8.8% to 16.6% fewer problem behaviors. The data are cross-sectional and mostly parent-reported, so they cannot show that screens or short sleep cause the behaviors, and the mediation effects were small.
The international journal of behavioral nutrition and physical activity
Among 487,200 Chinese adults followed for a median of 9.6 years, each of three insomnia symptoms was linked to higher cardiovascular risk, with adjusted hazard ratios from 1.07 to 1.13, and having all three went with an 18% higher risk of cardiovascular disease. Insomnia was self-reported once at baseline, so the link is an association, not proof of cause.
In 2737 US adults from NHANES 2013/2014, metabolic syndrome severity score followed a U-shaped curve with sleep duration (EDF = 2.94, P = 0.0004), lowest at 7 hours per night. The survey is cross-sectional, so it cannot show whether sleep length affects metabolic health, and the U-shape for metabolic syndrome itself was not statistically significant.
Among 11 872 UK adolescents aged 13-15, very high social media users (5+ hours a day) were more likely than average users to report late sleep onset (OR 2.14, 95% CI 1.83 to 2.50) and late wake times (OR 1.97) on school days. The data are cross-sectional and self-reported, so they cannot show that social media shortens sleep.
Pooling 18 observational studies, people with depression who had sleep disorders had higher odds of suicidal behaviour (OR 2.45, 95% CI 1.33 to 4.52), with nightmares (OR 4.47) and insomnia (OR 2.29) standing out. The certainty of the evidence was rated low to very low, so the link is suggestive and cannot show that sleep problems cause suicidal behaviour.
Using genetic variants as proxies, a Mendelian randomisation study found morning preference was linked to lower breast cancer risk (odds ratio 0.88 per category, 95% CI 0.82 to 0.93) and longer sleep to higher risk (1.19 per hour, 1.02 to 1.39). Insomnia results were inconsistent, and genetic proxies for sleep habits can be imperfect, so this is suggestive rather than proof.
Searching 2839 unique articles, this systematic review found only 9 studies, with 13 mediation analyses, testing what links sleep to pain intensity; mood, depression and anxiety, attention to pain, helplessness, stress, fatigue and physical activity were proposed mediators. Many studies used cross-sectional data with minimal adjustment for confounders, so the mechanisms remain unsettled.
Using genetic variants for sleep duration in 395,803 UK Biobank participants, each extra hour of sleep per day was linked to 1% slower reaction time (95% CI 0-2%; P = 0.008) and 3% more visual-memory errors (95% CI 0-6%; P = 0.05), with little evidence of an effect on dementia (OR 1.19, 95% CI 0.65-2.19). Genetic variants for sleep duration can have other effects, so these estimates are not guaranteed causal.
In 13 988 children followed from 2 to 16 years, active atopic dermatitis (AD) was associated with nearly 50% higher odds of more sleep-quality disturbances (adjusted OR 1.48; 95% CI, 1.33 to 1.66), while total sleep duration differed by only 2 minutes a day. AD and sleep were self-reported, so measurement error and shared causes cannot be ruled out.
Among 1936 adults in Catania, Italy, 1314 (67.9%) reported adequate sleep quality, and each one-point rise in the Mediterranean diet score was associated with 10% higher odds of adequate sleep. This is a single cross-sectional survey, and the link was seen only in normal-weight or overweight people, not in those with obesity.
Using circadian-rhythm features from wearable and smartphone data in 55 patients with mood disorders, a random forest model predicted mood state for the next 3 days with 65% accuracy (AUC 0.7). Accuracy was higher for episode types, but the sample is tiny and the abstract does not describe testing on separate patients.
Among 2195 adults aged 65 or older in a rural Chinese town, 33.8% had poor sleep quality, more often in women (39.2%) than men (26.3%). This was a one-time survey, so the links with sex and illness show association, not cause.
Pooling 65 studies with 1,550,524 participants and 86,201 cancer cases, neither short sleep (OR 1.01, 95% CI 0.97-1.05) nor long sleep (OR 1.02, 95% CI 0.97-1.07) was linked to higher cancer risk. Two subgroup findings (short sleep in Asians, long sleep and colorectal cancer) did not hold up in the dose-response analysis.
Across 9 randomised trials with 557 people with insomnia, exercise lowered Pittsburgh Sleep Quality Index scores by 2.87 points (95% CI 1.79 to 3.95) and Insomnia Severity Index scores by 3.22 points, but did not clearly change sleep efficiency (0.56% lower, 95% CI 3.42% lower to 2.31% higher). The evidence was rated low to very low quality for the two questionnaires, and most trials had a high or unclear risk of selection bias.
In 199 Canadian National Team athletes, the Athlete Sleep Screening Questionnaire agreed well with a sleep medicine physician's rating (Cohen's kappa = 0.84), with 81% sensitivity and 93% specificity in a randomized subset of 46. That subset is small, and the claim that sleep improved afterwards comes from a group without a comparison arm.
Pooling 26 studies of 51,623 participants, obstructive sleep apnea (OSA) was associated with resistant hypertension (pooled OR 2.842, 95% CI 1.703-3.980) and with essential hypertension, with odds rising from mild (OR 1.184) to severe OSA (OR 1.561). These are associations from observational studies, so the pooled odds ratios cannot show that OSA causes hypertension.