2 Avoid a slow death

One U.S. study that tracked participants without assigning them to groups found that current smokers have a life expectancy at least 10 years shorter than those who never smoked. Quitting before age 40 removes about 90% of this excess mortality risk. Quitting between ages 25–34 adds roughly 10 years to life expectancy; between 35–44, about 9 years; and between 45–54, about 6 years (the original reports these in life-years). Similar Chinese research from the 2010s shows that urban male smokers have 1.65 times the death risk of non-smokers (RR 1.65), and rural male smokers have 1.22 times. At 10 years after choosing to quit, this smoking-related risk nearly disappears. The same holds after a fracture. Pooling 19 observational studies that only tracked people without assigning groups, smokers’ odds of non-union (the bone failing to knit) were about 2.3 times those of non-smokers (OR 2.32, 95% CI 1.76–3.06, adjusted for other factors). The excess was also seen in fractures of the tibia (shin bone) and in open fractures. In a Swedish randomized trial, 50 people who joined a 6-week smoking cessation program after emergency fracture surgery had at least one postoperative complication in 20% of cases, versus 38% among the 55 who did not join. For scheduled surgery, you can start quitting 4 to 8 weeks before the operation, with weekly coaching and nicotine replacement. That lowers postoperative complications by about 58% (RR 0.42, 95% CI 0.27–0.65, 2 trials, 210 people).

In 2004, secondhand smoke caused 603,000 deaths globally — about 1.0% of all deaths worldwide, with children accounting for 28% of those fatalities. Pooled data from 57 studies show that people exposed to secondhand smoke have about 1.28 times the chance (odds ratio) of developing hypertension of non-exposed individuals, 1.39 times for heart disease, 1.50 times for myocardial infarction, and 1.36 times for stroke. The risk is greater when exposure occurs at home rather than outside.

A pooled analysis of 41 randomized trials involving 17,395 participants shows that the smoking cessation rate with varenicline is 2.32 times that with placebo (RR 2.32, 95% CI 2.15–2.51; high-certainty evidence). Varenicline also outperforms bupropion (RR 1.36, a 36% relative increase) and single-form nicotine replacement therapy (RR 1.25, a 25% relative increase). Across 133 trials with 64,640 participants, nicotine replacement products raise success rates by about 55% versus no medication at all (RR 1.55, 95% CI 1.49–1.61). Using both a patch and a fast-acting form such as gum or lozenges yields about a 27% relative improvement over using one form alone (RR 1.27, 95% CI 1.17–1.37; high-certainty evidence from 16 trials with 12,169 participants). Side effects: compared with placebo, people taking varenicline were about 2.6 times as likely to have nausea (RR 2.61) and about 37% more likely to have insomnia (RR 1.37), and more of them had vivid dreams. The review says nausea is the most common, mostly mild to moderate, and it generally fades over time. The label of one U.S. nicotine patch gives a three-step taper: 21 mg for 6 weeks, then 14 mg for 2 weeks, then 7 mg for 2 weeks, then stop.

In the UK, 697 smokers were randomly assigned to two groups. One group quit on a predetermined date; the other cut their cigarette use by 75% during the two weeks before that date. Both groups received nursing support and used nicotine replacement therapy around the quit date. At four weeks, 49.0% (95% CI 43.8–54.2) of the “quit-on-date” group remained smoke-free, versus 39.2% (34.0–44.4) of the “taper-first” group; the relative risk was 0.80 (0.66–0.93). By six months, 22.0% (18.0–26.6) of the quit-on-date group stayed quit, compared with 15.5% (12.0–19.7) of the taper-first group; the relative risk was 0.71 (0.46–0.91). People who preferred tapering from the start also had a lower success rate at four weeks than people who preferred to stop at once (38.3% vs. 52.2%).

Across 52 trials involving 19,488 participants, the combination of medication and behavioral support led to a 1.83-fold higher quit rate than standard care or brief advice (RR 1.83, 95% CI 1.68–1.98; high-quality evidence). For the 43 trials that recruited in medical settings, this advantage rose to 1.97-fold (95% CI 1.79–2.18). When looking at phone-based support alone: among people who called the helpline themselves, those receiving multiple follow-up calls had a 38% higher quit rate than those who only got informational materials or a single consultation (RR 1.38, 95% CI 1.19–1.61; 14 trials, 32,484 participants). For people who did not call the helpline but received scheduled phone support from researchers, the success rate was 25% higher than those without such support (RR 1.25, 95% CI 1.15–1.35; 65 trials, 41,233 participants). Both findings are supported by evidence of moderate certainty.

This up-to-date review includes 80 randomized trials involving 29,861 participants. Compared with nicotine replacement therapy, nicotine e-cigarettes raise the quit rate by about 61% (RR 1.61, 95% CI 1.23–2.12; high-certainty evidence based on 11 trials with 4,114 participants). In absolute terms, that translates to roughly 4 extra quitters per 100 users (range: 1–7). The rate of serious adverse events is similar between groups (risk difference 0.01, 95% CI −0.01–0.02; moderate-certainty evidence). Compared with behavioral support alone or no support at all, e-cigarettes improve quit rates by about 75% (RR 1.75, 95% CI 1.39–2.20; low-certainty evidence).

Two large U.S. studies followed large groups of people and only recorded, without assigning groups; together they included around 118,000 participants and documented 36,000 deaths. Those drinking two or more sugary drinks daily had a 21% higher risk of death compared to those consuming less than one per month (HR 1.21). For those drinking one to two per day, the risk rose by about 14% (HR 1.14). The European EPIC study followed 450,000 people across 10 countries under similar conditions; again only recorded, with no groups assigned. Compared to those drinking less than one sugary drink per month, daily consumers of two or more sugary drinks faced an 8% higher risk (HR 1.08, 95% CI 1.01–1.16). Those drinking two or more sugar-free drinks daily had a 26% higher risk (HR 1.26, 95% CI 1.16–1.35). When all types of soft drinks were considered together, the risk increase was about 17% (HR 1.17, 95% CI 1.11–1.22).

Data from 17 Asian studies involving 388,000 participants show that chewing betel nut raises overall mortality risk by about 21% relative to non-chewers (relative risk 1.21, P=0.02). The death figure counts 179,600 of those participants. The risk of diabetes rises by roughly 47% (1.47), while metabolic syndrome risk increases by about 51% (1.51). Betel nut itself is also a known risk factor for oral and esophageal cancer. Oral cancer: pooling 50 studies, on the Indian subcontinent chewing betel nut without tobacco carries about 2.56 times the risk of oral and oropharyngeal cancer compared with not chewing (95% CI 2.00–3.28). With tobacco added it is about 7.74 times (5.38–11.13). In Taiwan, chewing without tobacco carries about 10.98 times the risk (4.86–24.84). In India and Taiwan, the more you chew each day and the more years you chew, the higher the risk. Esophageal cancer: pooling 12 Asian studies, chewers have about 3.05 times the risk of esophageal squamous cell carcinoma (2.41–3.87). Those who both chew and smoke have about 6.79 times (4.71–9.79). After quitting: pooling 13 studies, people who quit betel nut with tobacco have about 34% lower risk of head and neck cancer than those who keep chewing (RR 0.66, 0.54–0.81). Those who quit betel nut without tobacco have about 27% lower risk (RR 0.73, 0.56–0.95). Those who have quit for 20 years or more have about 63% lower risk (RR 0.37, 0.22–0.61).

A randomized trial conducted in rural China enrolled 20,995 participants — all stroke survivors or adults aged 60+ with hypertension — and followed them for 4.74 years. Those using low-sodium salt had a 12% lower risk of death (RR 0.88), a 14% lower risk of stroke (RR 0.86), and a 13% lower risk of major cardiovascular events (RR 0.87) compared with the regular-salt group. No statistically significant difference emerged between the groups regarding episodes of excessively high blood potassium levels.

A Japanese study followed 9,676 participants for six years. Those who used interdental cleaning tools saw their mortality risk drop by about 11% (hazard ratio 0.89). Those who used tongue cleaners saw a 23% reduction (HR 0.77). When multiple studies of community-dwelling seniors are combined, people with no natural teeth have roughly 1.87 times the risk of death (95% CI 1.35–2.59). Those with fewer than 20 natural teeth face about 2.04 times the risk (95% CI 1.67–2.49). Electric versus manual: pooling 51 randomized trials, after 1 to 3 months of use electric brushes left about 11% less plaque, and after 3 months or more about 21% less. Gingivitis was about 6% and 11% milder respectively. The review authors say it is unclear how much these differences matter for health.

Fifteen studies that followed groups of people, recording only and not assigning groups, were pooled. These studies included 47,471 participants and 3,013 deaths. When grouped by daily step counts, the lowest group had a median of 3,553 steps/day; the next three groups had medians of 5,801, 7,842, and 10,901 steps/day. Their mortality risks were roughly 40%, 45%, and 53% lower than the lowest group (HR 0.60, 0.55, 0.47 respectively). For those 60 or older, benefits plateau after reaching 6,000–8,000 steps/day; for those 60 or younger, the plateau appears at 8,000–10,000 steps/day. Another pooled analysis found benefits begin at around 3,867 steps/day; each additional 1,000 steps/day reduces mortality risk by roughly 15%. By exercise time, pooling several follow-up studies (recording only, no groups), individuals meeting 1–2 times the recommended weekly activity level (7.5–15 MET·h/week, equivalent to 150–300 minutes of brisk walking) had a mortality risk about 31% lower than inactive people (HR 0.69). Those below this threshold still saw a 20% reduction (HR 0.80), while those exceeding it 3–5 times saw a maximum reduction of 39% (HR 0.61). Going higher brings no lower risk, but no harm either (HR 0.69 at 10 or more times the threshold). Another pooled study measured activity with accelerometers: the quarter of people with the most moderate-to-vigorous activity had a mortality risk about 48% lower than the quarter with the least (HR 0.52). An accelerometer is a device worn on the body that measures actual activity. This pooled study covered 8 studies and 36,383 people, with an average age of 62.6 years and a median follow-up of only 5.8 years. The authors removed those who died in the first two years and recalculated; the result barely changed, but they also say that bias from people moving less because they were already ill may remain.

Pooled data from multiple randomized controlled trials show that a 10 mmHg reduction in systolic blood pressure (the “high” number in a blood pressure reading) lowers mortality by about 13% (RR 0.87) and major cardiovascular events by roughly 20% (RR 0.80). Stroke risk drops by around 27% (RR 0.73), while heart failure risk decreases by about 28% (RR 0.72). Statins also reduce LDL cholesterol; each 1.0 mmol/L drop lowers mortality by roughly 10% (RR 0.90) and major vascular events by about 22% (RR 0.78). Observational studies further indicate that patients who adhere to prescribed treatment regimens (≥80% compliance) have a lower mortality risk than those who often miss doses: for statins, about 45% lower (RR 0.55); for blood pressure drugs, about 29% lower (RR 0.71). On how often to test blood lipids, the Healthy China Action advice is: people 40 or younger with normal lipids test every 2 to 5 years; people aged 40 and over test at least once a year; people at high risk of heart and brain vessel disease test every 6 months. The 2023 Chinese guideline on lipid management words it the same way around age 40, and for high-risk people it says to test as each person needs.

A pooled analysis of 16 longitudinal studies — involving 1.38 million participants and 113,000 deaths — shows that short sleepers face a 12% higher risk of death (RR 1.12), while long sleepers face a 30% higher risk (RR 1.30). When sleep duration is grouped into categories, the 7-hour bracket shows the lowest risk; each additional hour beyond 7 raises risk by about 13% (RR 1.13), and each hour less than 7 raises it by roughly 6% (RR 1.06). In the UK Biobank, wrist-watch data from about 61,000 participants were used to rank them into five groups based on schedule regularity; the four more regular groups had a 20%–48% lower mortality risk than the least regular group. How regular the schedule is predicts death better than how long you sleep. Another UK Biobank study used wrist-watch data on the timing of sleep in 51,562 people, of whom 3,853 developed cardiovascular disease during follow-up. Social jet lag means the gap between the midpoint of sleep on workdays and on weekends. Those with a social jet lag of 2 hours or more faced about a 30% higher risk of cardiovascular disease (HR 1.30, 95% CI 1.11–1.54), and this link persisted even among people getting a normal amount of sleep.

A UK-based study tracking 80,300 participants compared those who engaged in racket sports with those who did not. The former group had a 47% lower overall mortality risk (hazard ratio 0.53, 95% CI 0.40–0.69). Their risk of cardiovascular death was 56% lower (0.44, 95% CI 0.24–0.83). Racket sports here include tennis, badminton, and table tennis. For swimmers, these figures were 0.72 and 0.59 respectively; for aerobic exercise enthusiasts they were 0.73 and 0.64, and for cyclists 0.85. For running and soccer, no statistically significant difference was seen.

A study conducted by the UK Biobank involved 25,200 participants who did not engage in regular exercise. After an average follow-up period of 6.9 years, 852 participants passed away. Those who completed three 1–2 minute bouts of strenuous activity daily saw their overall mortality risk and risk of cancer-related death drop by 38%–40%. Their risk of cardiovascular death fell by 48%–49%. Even those who accumulated just 4.4 minutes of such activity per day experienced a 26%–30% reduction in overall mortality and cancer death risk, as well as a 32%–34% drop in cardiovascular death risk.

When multiple observational studies are pooled together (recording only, with no assigned groups), individuals who perform strength training show a 10%–17% reduction in mortality risk relative to non-trainers. This benefit peaks at roughly 30–60 minutes of weekly training, delivering a 10%–20% risk drop; more training does not further improve outcomes. Those who also do aerobic exercise enjoy an even lower mortality risk than those who do neither type of exercise.

One U.S. observational study used accelerometers to measure activity levels without grouping participants. It divided people into four groups based on their total daily sitting time. Those in the group with the most daily sitting time had 2.63 times the risk of death of those in the group with the least (HR 2.63). The study also split participants into four groups based on the longest continuous sitting period; those in the group with the longest continuous sitting time had 1.96 times the risk of those in the group with the shortest continuous sitting time (HR 1.96). This shows that both total daily sitting time and the length of each continuous sitting period are independently linked to mortality risk. Another pooled analysis combined studies of over a million people, again recording only and not assigning groups. It found that people who sat for 8 or more hours a day and did little to no exercise had 1.59 times the risk of death (HR 1.59) of those who sat for less than 4 hours a day and were the most active. About 60–75 minutes of moderate-intensity activity a day can cancel out the extra risk from sitting. For the most active group, even if they sat for 8 or more hours a day, their risk ratio was only 1.04 — no statistically significant difference was observed.

When multiple long-term studies are pooled together (without separating participants into groups), the group eating the most processed meat faces about a 23% higher risk of death (RR 1.23) than the group eating the least. For all red meat combined, the increase is about 29% (RR 1.29). Unprocessed red meat raises the risk by roughly 10% (RR 1.10), though this difference isn’t statistically significant. Another analysis looks at daily intake: each extra daily serving of processed meat adds about a 23% higher death risk (RR 1.23), while each extra serving of red meat adds about 10% (RR 1.10).

A combined analysis of 83 longitudinal studies involving 600,000 drinkers showed that the lowest mortality risk occurs at consumption levels of no more than 100 g of pure alcohol per week — roughly 2.5 L of beer at 5% alcohol content, or 300 mL of 40% alcohol spirits. At age 40, drinking 100–200 g per week shortens life expectancy by about 6 months; 200–350 g shortens it by 1–2 years; and over 350 g shortens it by 4–5 years. The Global Burden of Disease Study 2016 also concluded that zero alcohol consumption per week carries the lowest overall health risk. One adjusted analysis found that daily consumption of 1.3–24 g of alcohol was linked to a 7% lower mortality risk (RR 0.93) compared to lifelong abstainers, though this difference was not statistically significant. Daily consumption of 45–64 g raised risk by 19% (RR 1.19), and 65 g or more raised it by 35% (RR 1.35).

A 2026 clinical review notes that chronic heavy drinkers often develop withdrawal syndrome after abrupt cessation or a major reduction in intake; symptoms include tremors, autonomic hyperactivity, anxiety, insomnia, perceptual disturbances, seizures, and delirium tremens — conditions that are “common and potentially fatal” in general medical practice. Norway’s nationwide registry included 36,287 individuals diagnosed with alcohol dependence, withdrawal states, or delirium tremens between 2009 and 2015. The annual mortality rate was 8.0% in the delirium tremens group, 5.0% in the withdrawal state group, and 3.6% in the alcohol dependence group. The standardized mortality ratio for the delirium tremens group was 9.8 (95% CI 8.9–10.7), indicating that their risk of death is roughly 9.8 times higher than that of age-matched peers.

Data from 34 randomized trials involving 15,197 participants show that individuals receiving brief interventions were drinking 20 grams less pure alcohol per week after one year than those receiving minimal or no intervention (95% confidence interval: 12–28 grams; evidence of moderate quality). The average baseline consumption among these participants was 244 grams per week. “Brief interventions” here refer to no more than five counseling sessions totaling under 60 minutes; longer sessions provide no extra advantage. Regarding medications: 122 trials with 22,803 participants indicate that acamprosate helps one additional person per 12 treated to abstain from any drinking (95% CI: 8–26). With oral naltrexone at 50 mg daily, for every 12 people treated (8 to 26), one more person does not return to heavy drinking.

Two U.S.-based longitudinal studies tracked 119,000 participants over more than 3 million person-years. Compared to non-nut eaters: individuals consuming nuts less than once a week saw a 7% lower mortality risk (hazard ratio 0.93, 95% CI 0.90–0.96). Those eating nuts once a week had a 11% lower risk (0.89, 0.86–0.93). For intake of 2–4 times weekly, the reduction was 13% (0.87, 0.83–0.90); for 5–6 times weekly, 15% (0.85, 0.79–0.91). Eating them 7 or more times a week came with about a 20% lower risk (0.80, 0.73–0.86).

This conclusion comes from pooling data across six U.S. studies involving nearly 30,000 participants. Consuming two extra servings of processed meat weekly raised mortality risk by about 3% (hazard ratio 1.03, 95% CI 1.02–1.05). The same increase applied to unprocessed red meat (1.03, 1.01–1.05). However, replacing those servings with poultry resulted in a ratio of 0.99 (0.97–1.02), and fish gave 0.99 (0.97–1.01); neither showed any statistically significant difference.

Pooled data from multiple observational studies (no intervention groups used) shows that adding 90 g of whole grains each day lowers mortality risk by roughly 17% (RR 0.83). This 90 g figure corresponds to three servings. The protective effect remains evident up to 210–225 g per day. Another pooled analysis found that each additional daily serving reduces risk by about 8% (RR 0.92).

A study conducted in China involving 100,900 participants, called China-PAR, tracked half of them for 7.3 years. The results showed that people with a regular tea-drinking habit had a 15% lower risk of death (hazard ratio 0.85, 95% CI 0.79–0.90, which is a reliable range). For those starting at age 50, regular tea drinkers gain 1.41 extra years free from atherosclerotic cardiovascular diseases, and 1.26 extra years of overall life expectancy.

One review brought together 201 pooled analyses. The original studies under those pooled analyses only recorded what people did; they did not assign people to groups. On overall mortality, the review cited one of those pooled analyses of dose and effect. People who drank 3–4 cups of coffee a day, compared with people who drank none, had a 17% lower risk of death (relative risk 0.83, 95% CI 0.79–0.88, the confidence interval). At 7 cups a day, the reduction shrank to 10% (relative risk 0.90, 95% CI 0.85–0.96), and the risk was not seen to turn back up. A study from the U.S. National Health and Nutrition Examination Survey (NHANES) followed 40 725 adults. Half of them were followed for a full 9.8 years. It also only recorded; it did not assign groups. The study split coffee drinkers into two kinds. Morning-type drinkers took their coffee between 4:00 and 11:59 and hardly drank any in the afternoon or evening. They were 36%. Compared with people who drank no coffee, their risk of death from all causes was 12% lower (hazard ratio 0.88, 95% CI 0.81–0.96). After also taking out the effect of short sleep and trouble sleeping, it was 16% lower (hazard ratio 0.84, 95% CI 0.74–0.95). The risk of death from cardiovascular disease was 19% lower (hazard ratio 0.81, 95% CI 0.70–0.94). After taking out the effect of sleep, it was 31% lower (hazard ratio 0.69, 95% CI 0.55–0.87). All-day drinkers drank in the morning, the afternoon, and the evening. They were 16%. Their risk of death from all causes showed no difference (hazard ratio 0.99, 95% CI 0.90–1.10). Neither kind showed a difference in the risk of death from cancer. The daily caffeine ceiling has an official basis. Both the European Food Safety Authority (EFSA) and Health Canada set the ceiling for healthy adults at 400 mg a day. That is all sources added together. Tea, cola, and energy drinks count too. EFSA also says to spread it through the day, with no single occasion over 200 mg. One cup varies a lot. In EFSA's table, a 200 mL cup of filter coffee is about 90 mg. In Health Canada's table, an 8-ounce (237 mL) cup of brewed coffee averages 135 mg. In the U.S. Food and Drug Administration (FDA) table, 12 ounces (about 355 mL) of brewed coffee contains 113–247 mg. So the FDA turns 400 mg into about 2 to 3 cups of 12-ounce coffee, and the review turns it into about 4 cups.

When pooling data from multiple longitudinal studies (recording only, with no assigned groups), an additional 200 g of fruits and vegetables per day is linked to a 10% drop in death risk (RR 0.90). This protective effect persists up to 800 g per day. A combined analysis of two large US longitudinal studies and 26 other studies shows that people eating 5 servings daily have a mortality risk that is about 13% lower than those eating only 2 servings (HR 0.87). The optimal mix is 2 servings of fruit and 3 servings of vegetables; any extra amount beyond that does not lower the risk further.

One review brought together multiple pooled studies. The original studies under those pooled studies only recorded and did not assign people to groups. The results showed that people eating more ultra-processed foods had a 21% higher risk of death (RR 1.21) and a 50% higher risk of cardiovascular death (RR 1.50) compared to those eating less. These findings are classified as “highly suggestive” and “convincing” respectively. Yet when assessed using the standard GRADE criteria, the overall certainty of this evidence is low or very low.

A Chinese longitudinal study tracked 271,000 adults without cardiovascular disease, recording their health outcomes without assigning them to groups. Individuals who cooked using solid fuels such as coal or wood had a mortality risk roughly 11% higher (HR 1.11) than those using electricity or gas. Those who used solid fuels for heating faced a 14% higher risk (HR 1.14). Compared with people still burning solid fuels, people who had switched to cleaner fuels had about 13% lower risk for cooking (HR 0.87) and about 33% lower for heating (HR 0.67). Outdoor PM2.5 is a separate matter. Pooling 104 tracking studies, for every 10 µg/m³ increase in long-term exposure to outdoor PM2.5, the risk of death from natural causes rises by roughly 8% (RR 1.08). Natural causes refer to deaths not resulting from accidents.

In a high-risk area for esophageal cancer in northern Iran, researchers compared 300 patients with esophageal cancer to 571 healthy individuals. All results below are compared to people who drank lukewarm tea. Those who drank “hot” tea had roughly 2.07 times the risk of developing squamous cell carcinoma of the esophagus (95% CI: 1.28–3.35). For those who drank “very hot” tea, it was about 8.16 times (95% CI: 3.93–16.9). People who drank tea less than two minutes after pouring it had 5.41 times the risk of those who waited at least four minutes to drink it (95% CI: 2.63–11.1).

A Swedish study followed nearly 29,500 women for 20 years. Compared to the group getting the most sun, those who avoided it ended up living 0.6 to 2.1 years less. The authors noted that “non-smoking women who avoid sunlight have life expectancies comparable to smokers in the group getting the most sun exposure.”

This analysis combined individual data from 239 longitudinal studies (recording only, with no assigned groups). Only three groups remained in the final dataset: non-smokers, individuals without chronic illnesses at enrollment, and those who survived at least five years after enrollment. Results showed that mortality risk was lowest at BMI 20–25. Between 25–27.5, risk rose by about 7% (HR 1.07); at 27.5–30, it climbed by 20% (HR 1.20); at 30–35, by 45% (HR 1.45); at 35–40, by 94% (HR 1.94); and at 40–60, it was 2.76 times higher (HR 2.76). For East Asian populations, each additional 5 kg/m² of BMI, among those at BMI 25 or above, raised mortality risk by roughly 39% (HR 1.39).

A study conducted in Italy included 22,800 participants, with half of them followed for a full 8.2 years. Over this period, 1,236 participants passed away. Those who ate chili peppers four or more times per week had a 23% lower risk of death relative to people who ate very little chili (hazard ratio 0.77, 95% CI 0.66–0.90). Their risk of cardiovascular-related death was also 34% lower (hazard ratio 0.66, 95% CI 0.50–0.86).

The PURE study followed 136,000 participants across 21 countries for an average of 9.1 years; 6,796 of them passed away. Those who ate two or more servings of dairy products daily faced a 17% lower risk of death than non-consumers (hazard ratio 0.83, 95% CI 0.72–0.96). This trend was statistically significant (P=0.0052).

One U.S. follow-up study of 521,000 people recorded 129,000 deaths. Those who ate an extra half whole egg daily had a 7% higher mortality risk (hazard ratio 1.07, 95% CI 1.06–1.08, a reliable range). Replacing that half egg with equal amounts of egg whites, poultry, fish, dairy, or nuts and legumes further reduces mortality risk, including for cardiovascular disease, cancer, and respiratory conditions. The other side: three large U.S. studies with 216,000 people in total, followed for up to 32 years, found that people who ate at least one egg a day had no difference in cardiovascular disease risk compared with people who ate fewer than one a month (HR 0.93, 95% CI 0.82–1.05). The same paper pooled 33 sets of data on 1,720,000 people, and eating one more egg a day did not change cardiovascular risk either (RR 0.98, 0.93–1.03). The authors concluded that up to one egg a day is not linked to cardiovascular disease.

A Japanese study of 30,000 participants aged 40 to 59, followed for 19 years, identified 2,097 cardiovascular events. Those who bathed almost daily showed roughly 28% lower risk of cardiovascular events than those bathing no more than twice a week (hazard ratio 0.72, 95% CI 0.62–0.84). The risk of coronary heart disease was about 35% lower (0.65, 0.45–0.94), stroke risk dropped by roughly 26% (0.74, 0.62–0.87), and brain hemorrhage risk fell by about 46% (0.54, 0.40–0.73).

One umbrella meta-analysis compiled data from 16 separate meta-analyses covering 244 health outcomes. An umbrella meta-analysis essentially aggregates findings from multiple existing meta-analyses. Its conclusions: individuals who nap for less than 60 minutes show no rise in overall mortality or chronic disease risk, and experience the greatest improvement in cognitive function (SMD 0.69, 95% CI 0.37–1.00—a reliable range). Naps lasting 20–30 minutes provide the biggest boost to physical performance (SMD 0.99, 95% CI 0.67–1.31). On the other hand, people who nap for over 60 minutes face a roughly 30% higher risk of coronary heart disease, and about a 20% higher risk of diabetes and obesity. Another study used wrist-worn monitors to objectively measure naps in 1,338 adults aged 56 and older, rather than relying on self-reported questionnaires. After up to 19 years of follow-up, 926 participants had died. For every extra hour of napping, mortality risk rose by about 13% (HR 1.13, 95% CI 1.04–1.23). Each additional daily nap added roughly 7% to mortality risk (HR 1.07, 95% CI 1.02–1.13). People who napped in the morning faced a roughly 30% higher risk compared to those who napped earlier in the afternoon (HR 1.30, 95% CI 1.03–1.64).

This study involved 85,618 participants from the UK Biobank, with an average age of 61.8 years. Wrist-worn monitors tracked their daily sleep patterns over a median follow-up period of 8 years. The researchers first identified “sleep deficit nights” for each individual — nights when they got at least 2.5 hours less sleep than their personal sleep needs. The first night after such a deficit was labelled a “catch-up night”, and if participants got more sleep than required that night, it counted as a successful catch-up. Based on this, five distinct sleep patterns were identified. Compared to people with regular sleep schedules, those who were sleep-deprived and failed to catch up had a 15% higher all-cause mortality risk (HR 1.15, 95% CI 1.01–1.31). For those with a cumulative sleep deficit exceeding 3.5 hours and no catch-up, this risk rose to 42% (HR 1.42, 1.24–1.63). No significant risk elevation was found in either of the two groups that did make up their sleep deficit; for those who caught up after a moderate deficit, the HR was 1.12 (0.98–1.28). Among people who naturally get less sleep, those who were sleep-deprived and did not catch up had a 19% higher mortality risk (HR 1.19, 1.01–1.40), while those with severe sleep deficits and no catch-up faced a 38% higher risk (HR 1.38, 1.17–1.63). These findings were replicated in a separate US health survey involving 4,586 participants.

A pooled analysis of 23 studies — all observational, recording only, with no assigned groups — shows that night-shift workers face a roughly 13% higher risk of cardiovascular events (RR 1.13, 95% CI 1.10–1.16). Cardiovascular deaths rise by about 27% (RR 1.27, 1.18–1.36). In years, each extra 5 years of night work adds roughly 7% to cardiovascular disease risk (RR 1.07, 1.04–1.09) and about 5% to cardiovascular mortality (RR 1.05, 1.03–1.06). Specific conditions show similar trends: coronary heart disease incidence is about 22% higher (1.22, 1.16–1.28) and coronary death about 22% higher (1.22, 1.10–1.36); ischemic heart disease deaths are about 39% higher (1.39, 1.06–1.84); stroke deaths are about 49% higher (1.49, 1.04–2.12). Stroke incidence, however, shows no significant rise (RR 1.06, 0.95–1.18).

Aflatoxin is a Group 1 carcinogen per the International Agency for Research on Cancer, meaning it definitely causes cancer, especially liver cancer. A Guangzhou study measured aflatoxin B1 levels in homemade peanut oil from small producers between 2010 and 2022. The median concentration was 1.29 μg/kg, with half of all samples falling between 0.12 and 6.58 μg/kg. After local rules for small producers took effect, toxin levels dropped at once by 2.865 μg/kg, then fell an additional 2.593 μg/kg each year. Concurrently, the proportion with abnormal liver function among residents of the area where people often ate homemade oil fell by about 35% (PR 0.650, 95% CI 0.469–0.902). Another Guangxi study of 1,611 pregnant women found 81.7% ate homemade peanut oil; those women had a nearly 90% higher risk of bearing low birth weight babies (aOR 1.9, 95% CI 1.1–3.2) and an 80% higher risk of premature delivery (aOR 1.8, 95% CI 1.1–3.0).

A Cochrane review analyzed 15 randomized trials involving 56,675 participants, each followed for at least two years. Reducing saturated fat intake lowered the combined incidence of cardiovascular events by about 17% (RR 0.83, 0.70–0.98). Total mortality remained virtually unchanged (RR 0.96, 0.90–1.03) as did cardiovascular death (RR 0.95, 0.80–1.12). No significant difference was found between replacing saturated fats with polyunsaturated fats (the main component of soybean and corn oils) or with carbohydrates. Data on replacing them with monounsaturated fats (dominant in canola and olive oils) are too limited to draw conclusions. Another Cochrane review of 19 trials with 6,461 participants found that higher intake of omega-6 fatty acids (linoleic acid) showed no benefit or harm for overall mortality (RR 1.00, 0.88–1.12) or cardiovascular events (RR 0.97, 0.81–1.15). A third review of 86 trials involving 162,796 participants showed that increased consumption of plant-derived omega-3 fatty acids (found in flaxseed oil and perilla oil) had virtually no impact on overall mortality (RR 1.01, 0.84–1.20) or coronary events (RR 1.00, 0.82–1.22).

Ten case-control studies involving non-smoking Chinese women were pooled for analysis. In these studies, researchers compared lung cancer patients with healthy individuals, asking about past cooking habits. Overall, exposure to cooking fumes was linked to about 1.74 times the chance of lung cancer (odds ratio, 95% CI 1.57–1.94). Four of these studies specifically looked at women exposed to fumes who used no range hood; their chance was about 2.11 times (95% CI 1.54–2.89). One Taiwanese study compared 1,302 non-smoking women with lung cancer to 1,302 healthy women, focusing on those with at least ten years of cooking experience. Women who used their range hood for more than two-thirds of those years had roughly half the risk of developing lung cancer compared to those who used it for less than a third of the time (OR 0.49, 95% CI 0.32–0.76). A Shanghai-based prospective study followed 71,000 non-smoking women from 1996 to 2009, identifying 429 cases of lung cancer. Women who reported living in a poorly ventilated kitchen had a 49% higher risk of lung cancer (HR 1.49, 95% CI 1.15–1.95). Those who had burned coal in such kitchens for 20 years or more faced a 2.03-fold increase in risk (HR 2.03, 95% CI 1.35–3.05). No significant difference in risk was found between women who did and did not burn coal at all (HR 1.03, 95% CI 0.84–1.26), nor between those using different types of cooking oil.

A Cochrane review pooled 24 randomized trials with 3,329 participants, comparing oral hormones with placebo. Women on hormones had about 75% fewer hot flashes per week than on placebo (95% CI 64.3–82.3). Severity also fell clearly (OR 0.13, 0.07–0.23). Women on placebo also had about 57.7% fewer hot flashes, which is why anything claimed to treat hot flashes has to be compared against placebo to count. The risks come from the U.S. WHI randomized trial. It included 16,608 women aged 50 to 79 with a uterus, who took estrogen plus progestin and were followed for an average of 5.2 years. Coronary heart disease was about 29% higher (HR 1.29, 1.02–1.63). Invasive breast cancer was about 26% higher (HR 1.26, 1.00–1.59). Stroke was about 41% higher (HR 1.41, 1.07–1.85). Pulmonary embolism was about 2.13 times as high (HR 2.13, 1.39–3.25). In absolute terms, per 10,000 women per year there were 7 more cases of coronary heart disease, 8 more strokes, 8 more pulmonary emboli and 8 more invasive breast cancers, along with 6 fewer colorectal cancers and 5 fewer hip fractures. Among women who had had a hysterectomy and took estrogen alone, followed for a cumulative 13 years, breast cancer was actually about 21% lower (HR 0.79, 0.65–0.97). Pooling the two trials with 18 years of follow-up, overall mortality did not change (HR 0.99, 0.94–1.03). By age at starting, the 50-to-59 group had about 31% lower overall mortality during the years on treatment (HR 0.69, 0.51–0.94). The 60-to-69 group was 1.04 (0.87–1.25) and the 70-to-79 group 1.13 (0.94–1.36), neither significant. At 18 years of follow-up the 50-to-59 group was 0.89 (0.79–1.01), no longer significant. The 2017 Cochrane review, looking only at women aged 50 to 59, found the only clear increase was blood clots with combined estrogen-progestin therapy, with an absolute risk below 1/500. Among women aged 65 or older on combined therapy, dementia rose from 9 per 1,000 to 11–30 per 1,000 over 4 years. In 2022 the U.S. Preventive Services Task Force recommended against hormone therapy for preventing chronic conditions (grade D) and concluded there is no net benefit.

Pooling 10 observational follow-up studies covering about 3,840,000 people, followed for an average of 10.1 years, dog owners had about 24% lower death risk (RR 0.76, 95% CI 0.67–0.86). Cardiovascular death was about 31% lower (RR 0.69, 0.67–0.71). Among people who had had coronary heart disease, those with a dog at home were about 65% lower (RR 0.35, 0.17–0.69). The authors state that this pooled estimate was not adjusted for other factors. A Swedish nationwide register study of about 3,430,000 people adjusted for income, marital status and the like. For everyone combined, dog owners had about 20% lower death risk (HR 0.80, 0.79–0.82). Among people living alone, dog owners had about 33% lower death risk (HR 0.67, 0.65–0.69); among those living with others, about 11% lower (HR 0.89, 0.87–0.91). Among Swedish heart-attack patients, those living alone with a dog had about 33% lower death risk (HR 0.67, 0.61–0.75), and those living with a partner or children about 15% lower (HR 0.85, 0.80–0.90). After ischemic stroke the figures were about 27% lower (HR 0.73, 0.66–0.80) and about 12% lower (HR 0.88, 0.83–0.93) respectively. Counter-evidence: in a UK pooled analysis of 6 health surveys with 59,352 people, followed for an average of 11.5 years, dog ownership showed no link with death after full adjustment (HR 1.03, 0.98–1.09), nor with cardiovascular death (HR 1.07, 0.96–1.18). A Norwegian study with a median follow-up of 18.5 years found virtually the same death risk for dog owners (HR 1.00, 0.91–1.09), and dog owners there were no more active than others. In the Swedish twin register, among 34,202 people who could be asked about lifestyle, the link between dog ownership and lower cardiovascular disease and death risk could not be reproduced. An Australian study of 15,735 people followed for 4 years found dog owners’ death risk about 23% lower (OR 0.77, 0.59–0.99), part of which could be explained by more physical activity. Owning a cat, bird or fish showed no link with death. The American Heart Association’s 2013 statement says that pet ownership, especially dog ownership, “may be reasonable” for lowering cardiovascular risk, the weakest tier of recommendation. The same statement explicitly advises against adopting, rescuing or buying a pet mainly to lower cardiovascular risk.

The Daqing study was a randomized trial in Daqing, Heilongjiang, with groups assigned by clinic. 577 adults with impaired glucose tolerance (one form of prediabetes) were put into a control group or one of three intervention groups: diet only, exercise only, or both. The intervention lasted 6 years. Over those 6 years the cumulative share who got diabetes was 67.7% in the control group, 43.8% in the diet group, 41.1% in the exercise group and 46.0% in the diet-plus-exercise group. After adjusting for differences in body mass index and fasting blood sugar, the risk of diabetes in the three groups was about 31%, 46% and 42% lower. Lean and overweight people (the cut-off was a body mass index of 25) had similar reductions. At 23 years, the intervention groups had about 41% lower cardiovascular death (HR 0.59, 95% CI 0.36–0.96, the credible range), about 29% lower total death (HR 0.71, 0.51–0.99) and about 45% lower diabetes (HR 0.55, 0.40–0.76). At 30 years, the onset of diabetes in the intervention groups was pushed back by 3.96 years (by the median). Cardiovascular events were about 26% lower (HR 0.74, 0.59–0.92). Severe small-vessel complications of the eyes, kidneys and nerves were about 35% lower (HR 0.65, 0.45–0.95). Cardiovascular death was about 33% lower (HR 0.67, 0.48–0.94). Total death was about 26% lower (HR 0.74, 0.61–0.89). People lived 1.44 years longer on average (0.20–2.68). The U.S. DPP trial randomly assigned 3,234 people with high blood sugar. The lifestyle group's goal was to lose at least 7% of body weight and to be active at least 150 minutes a week. After an average of 2.8 years, the lifestyle group's risk of diabetes was 58% lower than the placebo group's (95% CI 48%–66%). The group taking metformin had 31% lower risk (17%–43%), which was worse than changing lifestyle. At an average of 15 years, the lifestyle group was still 27% lower (HR 0.73, 0.65–0.83), but the gap narrowed each year. At a median of 21 years, total death in the lifestyle group did not differ from the placebo group (HR 1.02, 0.81–1.28). Cardiovascular death did not differ either (HR 1.18, 0.77–1.81). The Finnish DPS trial randomly assigned 522 overweight people with impaired glucose tolerance. Over 4 years the cumulative share who got diabetes was 11% in the intervention group and 23% in the control group. During the trial the risk was 58% lower. After the one-to-one coaching stopped and follow-up went on, it was still 43% lower at a median of 7 years. At an average of 10.6 years, total death did not differ clearly (HR 0.57, 0.21–1.58). Cardiovascular disease did not differ either (HR 1.04, 0.72–1.51). The line given by the Healthy China Action is: a fasting blood sugar of 6.1 to 7.0 mmol/L (not including 7.0) counts as prediabetes. A blood sugar of 7.8 to 11.1 mmol/L (not including 11.1) 2 hours after a sugar drink also counts. It advises people who are overweight or obese to bring their body mass index down to or near 24, or to lose at least 7% of body weight. Total daily calories should be at least 400 to 500 kcal lower, and moderate-intensity activity should be at least 150 minutes a week. Check fasting blood sugar or the 2-hour blood sugar after a meal every six months.

Aflatoxin is a Group 1 carcinogen of the International Agency for Research on Cancer (it definitely causes cancer in people), and the cancer it causes is liver cancer. It is most common in peanuts and corn, and ordinary cooking such as stir-frying, boiling and deep-frying hardly breaks it down. A Shanghai study followed 18,244 middle-aged men from 1986 and kept one urine sample from each at entry. Over about 70,000 person-years, 55 liver cancers were found. 50 of them were compared with 267 people from the same group who did not get the disease. Those with aflatoxin found in their urine had about 3.4 times the risk of liver cancer (RR 3.4, 95% CI 1.1–10.0, the credible range). Those positive for hepatitis B surface antigen had about 7.3 times (2.2–24.4). Those with both had about 59.4 times (16.6–212.0), compared with people who had neither. In an earlier analysis of the same group, after adjusting for hepatitis B, smoking, drinking and education, those with the toxin found had about 3.8 times the risk (1.2–12.2). Qidong in Jiangsu used to be an area with a high rate of liver cancer, and from the 1980s the local staple was switched from corn to rice. The median of the toxin marker in residents' blood fell from 19.3 pg/mg in 1989 to undetectable in 2009 (below 0.5). Among residents 35 or younger, liver cancer death rates in the cohorts born from the 1960s to the 1980s fell by more than half. All of these cohorts were born before newborns were widely given the hepatitis B vaccine. The authors estimate that about 65% of the fall in liver cancer deaths can be put down to eating less of the toxin, and 83% of that fell among people infected with hepatitis B. By global estimates, a hepatitis B carrier who takes in the same amount of toxin has about 30 times the cancer-causing risk of someone without the virus. Each year in China a number of liver cancers can be put down to this toxin. About 5,300 to 14,400 of them occur in hepatitis B carriers, and about 1,990 to 4,430 in people without the virus.

Clonorchis sinensis is commonly called the liver fluke. In 2009 a meeting of the International Agency for Research on Cancer, part of the World Health Organization, assessed that long-term infection with it causes cancer in people. It causes bile duct cancer and is listed as Group 1. The 1994 assessment had given it only Group 2A (probably carcinogenic). Pooling published studies, people infected with the liver fluke have about 4.8 times the risk of bile duct cancer. In the original it is a relative risk of 4.8 (95% CI 2.8–8.4, the credible range). This figure also includes another liver fluke from Southeast Asia. In the same paper hepatitis B was 2.6 times and hepatitis C 1.8 times; the liver fluke was the strongest of them. Another paper pooled 36 studies. In case-control studies, bile duct cancer was about 4.37 times (OR 4.37, 2.84–6.72). In follow-up studies it was about 10.43 times (RR 10.43, 2.90–37.47). Cholangitis and cholecystitis were about 7.80 times (RR 7.80). Gallstones were about 2.42 times (RR 2.42). Hepatocellular carcinoma was about 4.69 times (OR 4.69). No significant link was seen with cirrhosis. The heavier the infection, the more of these diseases. In the national survey of 2014 to 2015, the infection rate was 0.47%, about 6 million people. Four provinces had a rate above 1%: Guangdong, Guangxi, Heilongjiang and Jilin. Estimated from 2016 to 2021 surveillance data, the national rate in 2020 was about 0.67%, about 9.46 million people. Of this, Guangxi was about 8.92% and Guangdong about 2.99%. In 2024, 15 departments including the National Disease Control and Prevention Administration issued a prevention and control plan. The plan takes the average infection rate at surveillance sites in 2018 to 2022 as the baseline: Guangxi 8.75%, Jilin 4.38%, Heilongjiang 3.16%, Guangdong 2.95%. In 2026 the China National Center for Food Safety Risk Assessment pointed out that seasonings such as alcohol, vinegar and wasabi cannot effectively kill liver flukes. Freezing must be at −20℃ or below for at least 7 days in a row. Heating must bring the center of the food to 70℃ or above for 10 minutes or more. The International Agency for Research on Cancer also states that pickled or smoked freshwater fish can still infect. For treatment, a 2022 pooled analysis covered 26 trials and 3,340 people. The praziquantel regimen recommended by the World Health Organization is 25 mg per kg of body weight each time, 3 times a day, for 2 days. The predicted cure rate for Clonorchis sinensis was 98.5% (85.4%–99.9%).