6 The anti-list

The PHS II trial followed 14,641 male doctors for a median of 11.2 years. Both groups — those taking multivitamins and those who didn’t — showed the same rate of major cardiovascular events. The hazard ratio was 1.01 (meaning no real difference), with a 95% confidence interval of 0.91–1.10. Overall mortality was also the same, with a hazard ratio of 0.94 (95% CI 0.88–1.02). In 2022, USPSTF evaluated nine randomized controlled trials involving 51,550 participants and found no association between multivitamins and total mortality; they assigned it a Grade I rating, meaning evidence is insufficient to draw any firm conclusions. It looks like a good deal because “one pill for all nutrients” sounds right, and it costs only a few tenths of a yuan a day.

The VITAL trial followed 25,871 participants for 5.3 years (median); those taking 1 gram of fish oil daily showed no difference in major cardiovascular events or mortality compared to non-users. For major cardiovascular events HR was 0.92 (HR is the hazard ratio; 1 means the two groups are the same). The 95% CI of 0.80–1.06 is the plausible range for that number. It crosses 1, which means no difference can be seen. All-cause mortality HR was 1.02 (0.90–1.15). The ASCEND trial, involving 15,480 diabetic patients, also taking 1 gram daily, followed for a mean of 7.4 years, also found no difference in severe vascular events (RR 0.97, 0.87–1.08) or all-cause mortality (RR 0.95, 0.86–1.05). It looks like a good deal because many people take “eating fish is healthy” to mean “fish oil capsules are beneficial.”

The VITAL trial enrolled 25,871 people who took 2000 IU daily and followed them for a median of 5.3 years. Cancer, cardiovascular disease, and death showed no difference. For cancers that spread, HR was 0.96 (HR is a hazard ratio; 1 means the two sides match; 95% CI 0.88–1.06 is the credible range). For major cardiovascular events, HR was 0.97 (0.85–1.12). For all-cause death, HR was 0.99 (0.87–1.12). The D-Health trial enrolled 21315 Australians aged 60 and above who took 60000 IU monthly and followed them for 5 years. That trial also showed no difference: all-cause death HR 1.04 (0.93–1.18). Death rates were 5.3% versus 5.1%. It looks worthwhile because some studies only record and never intervene, find that people with low vitamin D die more often, and that is misread as “supplementing works.” People who rarely go out have a separate recommendation. The UK Department of Health advises adults who rarely go out, live in care homes, or cover most of their skin when outdoors to take 10 micrograms of vitamin D daily year-round. The NHS states that 1 microgram equals 40 IU, so 10 micrograms is 400 IU. China’s health industry standard recommends 10 micrograms a day for adults aged 18 to 64 and 15 micrograms for those 65 and above. The same standard sets the tolerable upper intake at 50 micrograms a day (2000 IU), because too much raises blood calcium. The NHS adult upper limit is wider: 100 micrograms (4000 IU). Fatigue has many causes; the NHS lists insufficient sleep, stress and depression, iron-deficiency anemia, diabetes, hyperthyroidism, and sleep apnea. It advises seeing a doctor if fatigue lasts several weeks without a clear reason, or if it interferes with daily life. More is not better. VITAL also counted fractures: people on 2000 IU daily matched those who did not supplement (HR 0.98, 0.89–1.08). VITAL also looked at depression among 18353 of the participants aged 50 and above who were not depressed at the time. Those who supplemented had 12.9 cases of depression or relapse per 1000 person-years; those who did not had 13.3; no difference (HR 0.97, 0.87–1.09). Mood scores also did not differ. The authors state that the results do not support adults using vitamin D to prevent depression. A Canadian trial enrolled 311 healthy people aged 55 to 70 without osteoporosis, in three groups taking 400, 4000, or 10000 IU daily for 3 years. The higher the dose, the more bone density fell at the wrist radius: 1.2%, 2.4%, and 3.5% respectively. Bone compressive strength showed no difference. The authors conclude that this does not support high-dose vitamin D helping bone. A Swiss trial enrolled 200 people aged 70 and above who had fallen; they took 24000 IU or 60000 IU monthly for a year. The 60000 IU group’s leg function did not improve, and more of them fell within the year (66.9% versus 47.9%). An Australian trial enrolled 2256 women aged 70 and above who took 500000 IU once each autumn–winter. Their falls were about 15% higher (RR 1.15, 1.02–1.30) and fractures about 26% higher (RR 1.26, 1.00–1.59).

Cochrane reviewed 78 randomized controlled trials with 296,707 participants total. The overall risk ratio for mortality among supplement users was 1.02 (95% CI 0.98–1.05). When only higher-quality trials were considered, the risk ratio rose to 1.04 (1.01–1.07). Specifically, beta-carotene raised mortality risk to 1.05 (1.01–1.09) and vitamin E to 1.03 (1.00–1.05). The ATBC trial tracked 29,133 male smokers who took 20 mg daily; it found a 18% increase in lung cancer incidence (3%–36%) and an 8% rise in overall mortality (1%–16%). The CARET trial involved 18,314 smokers or asbestos-exposed individuals; it showed an about 28% higher risk of lung cancer (RR 1.28, 1.04–1.57) and an about 17% higher risk of all-cause mortality (RR 1.17, 1.03–1.33). It looks like a good deal because of the chain of reasoning “antioxidants fight aging, and fighting aging prevents cancer.” Each step sounds like common sense.

The GAIT trial enrolled 1583 people with knee osteoarthritis for 24 weeks. Among those on placebo (fake medicine), 60.1% also said pain eased. Glucosamine was only 3.9 percentage points higher (P=0.30). Chondroitin was 5.3 percentage points higher (P=0.17). Both together were 6.5 percentage points higher (P=0.09). These gaps are too small to rule out chance. In the same trial the real control drug celecoxib was 10.0 percentage points higher (P=0.008). It looks worthwhile because “cartilage building blocks” sound on-target, and placebo alone already helped six in ten people say it worked.

Cochrane pooled 29 controlled trials involving 11,306 participants. For ordinary people, regular vitamin C consumption does not meaningfully reduce overall cold incidence; the risk ratio here is 0.97, meaning there is virtually no difference between groups. The 95% confidence interval for this figure is 0.94–1.00. Trials where vitamin C was taken only after cold symptoms appeared showed mixed results: some indicated a benefit, others did not. In the same review, long-term daily intake shortens cold duration by about 8% in adults (95% CI 3%–12%) and 14% in children (95% CI 7%–21%). Marathon runners, skiers, and soldiers in cold climates again form an exception: 5 trials with 598 participants showed their cold incidence dropped by roughly half, with a risk ratio of 0.48 (95% CI 0.35–0.64). The perceived value of vitamin C for cold prevention is largely driven by its low cost, pleasant sweet-and-sour taste, and the fact that this claim has been widely circulated for half a century.

The USPSTF, which evaluates preventive measures in the U.S., gave a D rating to using CA-125 and ultrasound to screen asymptomatic women for ovarian cancer — meaning it does not recommend this screening. Their basis is the PLCO trial, which showed a 0.34% death rate from ovarian cancer in the screened group versus 0.29% in the control group (RR 1.18, 95% CI 0.82–1.71). Screening brought no benefit. The rate of unnecessary surgery due to false positives ranges from 0.2% to 3.25% across studies, with up to 15% of those patients experiencing serious complications. Another study screened 1,192 symptom-free individuals with whole-body CT; 86% had at least one “abnormality” noted, and 37% were advised to undergo further tests. The paper says “most findings were described as benign.” It looks like a good deal because “check the whole body at once, find it early, treat it early” fits intuition so well.

The IDEA trial involved 471 overweight or obese adults and followed them for 24 months. Everyone followed identical diet and exercise routines; the only difference was whether they wore wearable devices. The group wearing trackers lost 3.5 kg (95% CI 2.6–4.5). The group not wearing them lost 5.9 kg (5.0–6.8), which is more. The difference was 2.4 kg (95% CI 1.0–3.7, P=0.002). It looks like a good deal because the idea “you must measure to manage” sounds logical, and the devices are a one-time purchase.

One systematic review combined findings from 17 human studies and 223 studies analyzing food composition. Its conclusion was that “there is insufficient evidence to show organic food is significantly more nutritious than conventional food.” Of those, just 3 studies looked at actual health outcomes (allergies and symptomatic infections); no differences were seen between the two groups. Organic produce had pesticide residues detected 30 percentage points less often, but a detection is not the same as exceeding the limit. It looks worthwhile because “no pesticides means safer, and safer means healthier,” and each step sounds like common sense.

What you get back is the money saved. The reasons are in the first 6 items of this section. The most common ingredients in health products — multivitamins, fish oil, vitamin D, antioxidants, glucosamine, vitamin C — showed no expected benefit in randomized controlled trials. The numbers for each are in those 6 items’ Benefit lines. Regulatory rules only say what may be claimed; they are not evidence that a product is useless. The State Administration for Market Regulation requires that the front of a health-food pack dedicate no less than 20% of its area to one sentence: “Health food is not a drug and cannot replace drug treatment of disease.” The same rules lock the use of health foods to “supplementing dietary nutrients, maintaining or improving the body’s health state, or lowering disease risk factors.” It looks worthwhile because “does not treat disease, only tones” is hard to falsify, and when the money is spent as a gift it rides on reciprocity, not evidence.

One review pooled 45 trials in healthy adults. Its conclusion: the trials “failed to support” probiotics causing lasting changes in gut flora, and also failed to show improved blood lipids in healthy adults. Gut bacteria do shift, but only temporarily; they go back once use stops. It looks worthwhile because “gut health” is a broad enough phrase to fit almost anything, and the daily price of one bottle is low.

A Dutch randomized controlled trial enrolled 3018 people who, after a hot shower, rinsed with cold water for 30, 60, or 90 seconds. Their sick-leave occasions fell by 29% (IRR 0.71, a rate ratio for counts). But the paper states there was “no significant between-group difference in sick-leave days”: total days ill were the same on both sides. Quality of life, work productivity, and anxiety also showed no meaningful differences. A 2025 systematic review pooled 11 randomized trials and 3177 people. Right after cold water and 1 hour later, markers of inflammation rose instead. Immune markers did not change. Stress fell only at the 12-hour time point. The authors themselves say the evidence is “limited by few RCTs and small samples,” meaning such randomized trials are scarce and each enrolls few people. It looks worthwhile because it costs nothing and the intuition “suffering works” is smooth.

One review concluded that “despite the booming detox industry, there is very little clinical evidence supporting these diets,” and that “no randomized controlled trials have been done on commercial detox diets.” Another systematic review examined diet pH, alkaline water, and cancer: after screening 8,278 studies, just one qualified for inclusion, and it showed no link either. The authors wrote, “It is not reasonable to promote alkaline diets and alkaline water to the public for cancer prevention or treatment.” These products look like a good deal because the assumptions “the body contains toxins needing removal” and “an acidic body state causes cancer” sound plausible from a physiological standpoint.

One literature review specifically checked the “8×8” claim — 8 glasses a day, 8 ounces each. Its conclusion was “no scientific study supporting 8×8 was found.” It also wrote that “surveys of food and fluid intake in thousands of adults strongly suggest such a large amount is not needed,” because food and other drinks already supply a lot of water. The author states clearly that this conclusion applies only to “healthy adults in a temperate climate who sit most of the time.” He also wrote that some treatments or prevention need 8×8 or more, and that “hard labor and exercise, especially in hot climates,” of course need more. The claim looks worthwhile because water is free, 8 is a neat number, and “drink more water” sounds harmless in any setting.

Classic classroom experiment: after a personality test, students got a “personal analysis” that was identical for everyone, yet most rated it as accurate. The author’s conclusion is that “acceptance by the subject or the analyst does not prove the explanation is correct.” Of 39 students scoring accuracy from 0 to 5, 34 gave 4 or 5. From the table distribution, the mean is about 4.26. That explains why fortune-telling always seems accurate — the Barnum effect. Astrology was also tested double-blind; the result was published in Nature. Astrologers had to pick, from three personality profiles, the one that matched a chart. They picked correctly in 40 of 116. Blind guessing would hit about 38.5; the astrologers had expected to hit 58.5. Official case: the Xixian County procuratorate in Henan reported an online fortune-telling fraud ring that first drew people in with free palm reading, then charged “incense money” and “offering packages” up to 3888 yuan. Over 50 victims; more than 2,000,000 yuan (200-plus wan) recovered; the lead defendant got 11 years’ imprisonment and a fine of 100,000 yuan.

A 2023 Cochrane systematic review included 17 randomized controlled trials, each with 5 to 156 people, followed from under 1 day to 5 weeks. Because data were sparse, the people enrolled differed across trials, and follow-up times did not match, the authors did not pool them into one total. Results below are versus lenses without a blue-light filter. At under one week of follow-up, self-reported eye fatigue “may not differ”; certainty is low. One trial of 120 people had a between-group difference of 9.76 units. The credible range for that difference is 95% CI −33.95 to 53.47. The range crosses positive and negative, so no difference shows. Critical flicker fusion frequency, an objective measure of eye fatigue, “may have almost no difference”; certainty low. Best visual acuity after refractive correction is “probably unaffected.” The difference was 0.00 logMAR (95% CI −0.02 to 0.02). That finding rests on 1 study and 156 people; certainty moderate. Effect on daytime alertness is unclear (2 trials, 42 people, certainty very low). The review also found no evidence supporting protection of the macular retina. An American Academy of Ophthalmology public page states that the claim “long screen time permanently damages the eyes” “is not true”; screens only make the eyes uncomfortable after long use. China’s Classification and Catalogue of Occupational Diseases (2024 edition) lists photic eye injury from welding arcs and laser eye injury as occupational diseases.

The Finnish Kuopio study followed 2862 men aged 42 to 60 for nearly 9 years on average. Of 153 who had donated in the 24 months before counting started, 1 (0.7%) had an acute heart attack. Of 2529 who had not donated, 316 (12.5%) did (difference between the two rates P<0.0001). After subtracting age and other coronary risk factors, donors’ heart-attack risk was about 88% lower. Relative risk 0.12 (95% CI 0.02–0.86, P=0.035). But that is only from follow-up records, with no randomized comparison. The Nordic SCANDAT database has 1182495 donors followed for 9526627 person-years, with 15401 deaths. Subtracting only traits such as age and sex, one extra donation a year linked to 18.6% lower mortality (95% CI 16.8%–20.4%). Subtracting also the layer “only the healthy stay allowed to donate,” the drop is left at 7.5% (5.7%–9.4%). The authors conclude it “cannot be interpreted as solid evidence of a health benefit.” The FeAST randomized controlled trial enrolled 1277 people with symptomatic lower-limb arterial disease, randomized into two groups: one had blood drawn every 6 months to lower iron; one was a control. The main outcome was all-cause death: 125/636 (20%) versus 148/641 (23%). HR 0.85 (HR is a hazard ratio; 1 means the two sides match). 95% CI 0.67–1.08, P=0.17. Death plus non-fatal heart attack and stroke together: 180 (28%) versus 205 (32%), HR 0.88 (0.72–1.07, P=0.20). Neither showed a difference. The INTERVAL randomized trial enrolled 45263 whole-blood donors in three donation intervals: men 12, 10, or 8 weeks; women 16, 14, or 12 weeks. Shorter intervals clearly collected more blood over two years (men’s 8-week group 1.69 units more than the 12-week group, 95% CI 1.59–1.80). Quality of life, physical activity, and cognition showed no between-group differences. But donation-related discomfort rose, including fatigue, shortness of breath, vertigo, dizziness, and restless legs (legs so uncomfortable at night you cannot lie still). Mean hemoglobin and ferritin were also lower, and delays for low hemoglobin were more frequent (all P<0.0001). Iron deficiency affects heat production: 10 women with iron-deficiency anemia, 8 with exhausted iron stores but no anemia yet, and 12 control women sat in 28 °C water for 100 minutes. The anemia group’s rectal temperature was 36.0±0.2 °C versus 36.2±0.1 °C in controls (P=0.001). Oxygen use was 5.28±0.26 versus 5.99±0.29 mL·min⁻¹·kg⁻¹ (P=0.04). Thyroxine and triiodothyronine were also lower (P<0.002). The claim looks true because donors are already repeatedly screened healthy people, and the hypothesis “removing iron protects the heart” did not show up in randomized trials.

The 2019 update of a Cochrane systematic review included 17 randomized controlled trials. For overall mortality it pooled 11 of them, 233298 participants, and 21535 deaths. The general-check group matched the control group: RR 1.00. RR is a risk ratio; 1 means the two sides match. 95% CI 0.97–1.03 is the credible range. For cardiovascular death it pooled 9 trials, 170,227 people, and 6237 deaths: RR 1.05 (95% CI 0.94–1.16). For cancer death it pooled 8 trials, 139,290 people, and 3663 deaths: RR 1.01 (95% CI 0.92–1.12). Neither showed a difference. The authors conclude that “general health checks are unlikely to be beneficial.” The national basic public-health service measures blood pressure free once a year for permanent residents aged 35 and above. It also gives people aged 65 and above one free health check a year. That free check covers physical exam, blood count, urine routine, fasting glucose, and ECG. Liver function: AST, ALT, total bilirubin. Kidney function: serum creatinine, blood urea. Lipids: total cholesterol, triglycerides, LDL cholesterol, HDL cholesterol. Plus abdominal ultrasound of liver, gallbladder, pancreas, and spleen (nationwide).

The American College of Rheumatology 2020 gout guideline covers asymptomatic hyperuricemia: blood uric acid >6.8 mg/dL, with no prior gout attack and no subcutaneous tophi. The guideline conditionally recommends that these people not start any uric-acid–lowering drug. Drugs include allopurinol, febuxostat, and probenecid. Evidence certainty is rated “high.” A table footnote states that “by attributable risk, 24 patients must be treated for 3 years to prevent one gout attack.” The CKD-FIX randomized controlled trial enrolled 363 people with stage 3 or 4 chronic kidney disease, no gout history, but risk of worsening. One group took 100–300 mg allopurinol daily; one took placebo. The trial ran 104 weeks. eGFR is a kidney-function marker in mL/min/1.73 m². One group changed −3.33 per year (95% CI −4.11 to −2.55). The other −3.23 (−3.98 to −2.47). The difference was −0.10 (−1.18 to 0.97), P=0.85 — kidney function fell at the same speed. Serious adverse events 46% versus 44%. The PERL randomized controlled trial enrolled 530 people with type 1 diabetes and early-to-mid diabetic kidney disease; they took allopurinol for 3 years, then stopped for 2 months to wash out the effect. Allopurinol did lower blood uric acid from 6.1 to 3.9 mg/dL. After washout, the two groups’ GFR differed by only 0.001 mL/min/1.73 m². The credible range for that difference is 95% CI −1.9 to 1.9, P=0.99. Urinary albumin was 40% higher on allopurinol (0 to 80). It looks worthwhile because the intuition “the marker is high so take a drug to push it down” is smooth.

Italy’s MICOL population survey ultrasound-scanned 11,229 people aged 29 to 69. It found 856 with gallstones, of whom 580 had no symptoms, followed a mean 8.7 years. Of those 580, 453 (78.1%) stayed symptom-free, 61 (10.5%) developed mild symptoms, and 66 (11.4%) severe symptoms. Over follow-up there were 189 cholecystectomies, 41.3% of them in people without symptoms. The authors conclude that “expectant management remains an effective approach for most patients” — watch first, no surgery. A U.S. HMO (a managed-care organization) followed 691 people with confirmed gallstones a mean 78±61.6 months. Of 135 without symptoms followed 58±50.2 months (median 46.3 months), only 10% developed gallstone symptoms and 7% had surgery. Of 50 deaths in the whole group, only 2 were biliary-related, both in the symptomatic group. The authors conclude that “patients with silent stones need not be operated on before symptoms appear”; silent stones are stones that stay quiet and do not flare. UK NICE guideline CG188 item 1.2.1 says: for people with a normal gallbladder and biliary tree who have incidentally found asymptomatic gallstones, “reassure them that they do not need treatment until symptoms develop.” The European Association for the Study of the Liver 2016 gallstone guideline puts both sides’ risks together. Asymptomatic people develop symptoms in about 0.7% to 2.5% per year. Acute cholecystitis, acute pancreatitis, obstructive jaundice (blocked bile duct, person turns yellow), or cholangitis run 0.1% to 0.3% per year. After a first colic, complications rise to 1% to 3% per year, so most people hurt once before a major event. On the surgery side, Sweden’s national registry of 51,041 cholecystectomies had bile-duct injury in 1.5%. About one fifth of those (0.3%) were partial or complete duct transection. One-year mortality was 3.9% with duct injury versus 1.1% without. The guideline also notes that after surgery 10% to 40% of patients still have discomfort, or it improves then returns. The guideline’s conclusion: operating on asymptomatic people, open or laparoscopic, does not lengthen life, because the death and complication risk of surgery exceeds the risk of not cutting. Waiting for symptoms also costs less overall. A 1983 estimate gave numbers: prophylactic removal in a 30-year-old man meant living 4 days less on average than watching first; at age 50, 18 days less. No randomized trial has yet compared cutting versus not. Prophylactic removal looks worthwhile because of the intuition “it will hurt eventually, better cut once while young.”

A 5-year randomized controlled trial enrolled 120 men with recurrent calcium-oxalate stones and high urine calcium (idiopathic hypercalciuria). They were randomized into two groups. One ate normal calcium, 30 mmol a day, while also cutting animal protein (52 g a day) and salt (50 mmol sodium chloride a day). The other ate a traditional low-calcium diet, 10 mmol a day. Over 5 years recurrences were 12/60 versus 23/60. Unadjusted relative risk 0.49 (95% CI 0.24–0.98, the credible range), P=0.04 — recurrence risk on normal calcium was half. During follow-up, urine calcium fell about 170 mg a day in both groups. But urine oxalate differed. The low-calcium group rose a mean 5.4 mg a day; the normal-calcium group fell a mean 7.2 mg a day. UK NICE guideline NG118 item 1.8.1 states not to restrict daily calcium intake. Adults keep 700 to 1200 mg a day. Children and teens 350 to 1000 mg a day by age. Giving up calcium looks worthwhile because of the intuition “stones are made of calcium, so eat less calcium.”

A classic classroom experiment showed that everyone got the same vague “personal analysis” yet most think it is accurate — the Barnum effect. It explains where the experience “things really went smoother after I put it on” comes from. That author’s conclusion is “acceptance by the person does not prove the explanation is correct.” Official case: on 4 June 2026 the State Administration for Market Regulation published typical “fake-fighting, source-clearing” cases in the traditional-craft market. A jewellery shop in Sihui, Guangdong, invented product materials on livestream. It also falsely claimed effects such as “preventing and treating rheumatism, raising immunity, improving sleep, and clearing blood vessels.” The shop used scripted acting to mislead and deceive consumers. That violates Article 9 Paragraph 1 of the Anti-Unfair Competition Law (nationwide).

One meta-analysis (combining findings from many studies) examined 259 separate samples and 753 effect sizes. The results showed that individuals who prioritized material possessions reported lower happiness levels. On the most commonly used multi-dimensional scale, the correlation coefficient was r = -0.19 (adjusted ρ = -0.24). On scales that measured only “valuing money,” the link was even weaker, with r values ranging from -0.08 to -0.11. The strength of this association varied depending on the outcome measured: it was strongest with risky health and spending behaviors and negative self-evaluations (r = -0.28 to -0.44), and weakest with life satisfaction and negative emotions (r = -0.13 to -0.15). Mediation analyses suggested this negative link may arise because unmet psychological needs drive it. Another study found that in two surveys of different populations, respondents both times said money spent on a life experience made them happier than buying things. In later lab experiments, good mood after recalling money spent on an experience was also higher than after recalling a purchase of a thing. Both studies used samples from outside China.

The U.S. National Survey of Families and Households (NSFH) revisited the same people in two waves. The authors matched it to local average income. Local average income (PUMA) was estimated from a 5% sample of the 1990 census. After subtracting own income and a set of personal traits, the result was: higher neighbour income, lower self-reported happiness. The effect is large, stable under other calculations, and highly significant. The authors’ phrasing is that the happiness drop from “neighbour income rising” is on the same order as from “own income falling by the same amount.” Comparing the same person to themselves over time (individual fixed effects), the coefficient’s direction stays and its size is similar; significance only falls to the 10% level. Restricting comparisons to the same state and the same survey wave (state and wave fixed effects), results barely change. So it is not that “people who like moving to poor places are happier anyway,” nor unmeasured state-level factors. The effect is stronger in people who often deal with neighbours and not significant in people who only deal with friends outside the community. Renters and owners did not differ, which does not support “expensive areas have high house prices so spendable money is less.” Happiness falls only with income of same-education neighbours; income of differently educated neighbours barely registers. And it mainly travels through “satisfaction with one’s own material situation,” not through satisfaction with family life or other domains. Finally, if own and neighbour income rise by the same proportion, happiness still rises.

A multi-lab preregistered replication had 23 laboratories and 2141 participants. They redid the sequential-task paradigm under Sripada and colleagues’ standardized protocol: first a depleting self-control task, then a second task that needs self-control. Pooled results showed a very small ego-depletion effect, and the 95% credible interval included 0 (d = 0.04, 95% CI −0.07 to 0.15). The authors discuss whether the “self-control is a limited resource” model still stands. Another similarly preregistered redo, 36 labs and 3531 people, found effect size d = 0.06 — again too small to count.

First the claim “skipping breakfast makes people eat more later.” A meta-analysis pooled 13 randomized trials. People assigned to eat breakfast took in 259.79 kcal more per day than those who skipped (95% CI 78.87–440.71, the credible range). Weight was 0.44 kg higher (0.07–0.82). Skippers did not make it up at lunch and dinner. Follow-up in these trials was short: mean 7 weeks for weight, mean 2 weeks for intake. Next, 16:8 and similar eating only in fixed hours. One trial put 116 overweight adults on it for 12 weeks. Eating only from noon to 8 p.m. versus three regular meals: weight differed by −0.26 kg (−1.30 to 0.78), not significant. Another in Guangzhou, 139 people with obesity, followed 12 months. Time-restricted eating plus calorie control lost 1.8 kg more than calorie control alone (−4.0 to 0.4), also not significant. Finally, whether long-term breakfast skipping harms. In a pool of 2.38 million people, skippers had about 17% higher cardiovascular-disease risk (OR 1.17, 1.09–1.26) and about 49% higher cardiovascular death (OR 1.49, 1.20–1.84). But these are record-only, non-randomized studies. In the original study followed 17 to 23 years, the all-cause-death hazard ratio was 1.19, credible range 0.99 to 1.42, crossing 1 — not a difference.

First the body. A U.S. study followed 31925 men through 2010. Each reported how many ejaculations a month, then was grouped by count. Versus 4 to 7 times a month, those with 21 or more had prostate-cancer HR 0.81 (95% CI 0.72–0.92, about 19% lower) using counts at ages 20 to 29. Using ages 40 to 49, HR 0.78 (0.69–0.89, about 22% lower). What fell was mainly low-risk prostate cancer. The opposing side is a UK case-control study comparing 431 patients diagnosed before 60 with 409 controls. Frequent masturbation in the twenties and thirties linked to higher risk; frequent in the fifties linked to lower risk. Next, porn. A systematic review plus meta-analysis holds that many people who feel “addicted” are mostly mismatched with what they believe is right, not only that they watch a lot. A few people do feel unwell after every ejaculation, but that is not “harming the body” — it is a rare disease called post-orgasmic illness syndrome (POIS). See a doctor; do not rely on abstinence. Among 45 Dutch patients, 87% began within 30 minutes of ejaculation, 80% were exhausted and weak, 78% felt like they had a cold. At a French hospital, 37 patients took a mean 4.7 days to recover. Cause is unclear. In the Dutch study, of 33 who had skin tests, 29 (88%) were allergic to their own semen. A Chinese study of 24 patients found no semen-specific allergy antibodies; anxiety and depression scores were high. A 2025 review found only 14 treatment studies, 12 of them case reports. Of 211 sexual-medicine experts worldwide, only 18% said more than three in ten patients improve (U.S. / UK / Netherlands / France / China).

The American Psychological Association’s 2009 task-force report says efforts to change sexual orientation are “unlikely to be successful and involve some risk of harm.” For sending minors into coercive or residential conversion programs, the report advises that neither parents nor children consider it. A U.S. nationally representative survey of 1518 sexual-minority adults found about 7% had undergone conversion, about eight in ten of those by religious practitioners. After subtracting age, adverse childhood experiences, and similar factors, those who had undergone conversion had lifetime suicidal ideation OR 1.92 (95% CI 1.01–3.64, about 90% higher). Planned suicide OR 1.75 (1.01–3.06). Suicide attempts with no or mild injury OR 1.88 (1.01–3.50). Another study interviewed 245 young people aged 21 to 25. Of those not converted in adolescence, 22.0% had attempted suicide. When parents themselves tried to change the child’s orientation, 48.1%. When parents also sent the child to a therapist or religious figure, 62.8%. WHO’s disease classification ICD-10 states: “Sexual orientation by itself is not to be regarded as a disorder.” In 2017 a court in Yicheng District, Zhumadian, Henan, heard a case: a psychiatric hospital forcibly admitted a man under “sexual preference disorder.” The court held that his “condition was ordinary, with no suicide, assault, destruction of property, or similar behaviour or danger, and he did not meet the conditions for compulsory treatment.” The hospital was ordered to apologize publicly and pay 5000 yuan in mental-distress damages (U.S. / China).

First training. The 2011 Cochrane systematic review pooled 9 randomized trials and 20101 employees. In 7 trials and 19,317 people: those who had lifting training reported about as much later low-back pain as those who did nothing (OR 1.17, 95% CI 0.68–2.02, credible range crosses 1). Versus watching a video only, RR 0.93 (0.69–1.25), also similar. Evidence quality medium. The authors conclude that lifting advice and training, with or without handling aids, do not prevent low-back pain or disability from it. Next, bending. A 2020 systematic review of how much the back bends when lifting rated evidence quality low. In the only follow-up study, people who later had low-back pain and those who did not differed by only 1.5 degrees in maximum bend while lifting (−0.7 to 3.7), not significant. Of 11 studies, 9 saw no between-group difference. Another 7 studies used a different measure: people with low-back pain bent 6.0 degrees less while lifting (−11.2 to −0.9). The authors conclude there is no evidence that bending more invites low-back pain. Finally what to do. One meta-analysis pooled 8 follow-up studies: each extra 10 kg lifted per time, low-back-pain risk OR 1.11 (1.05–1.18, about 11% higher). Each extra 10 lifts a day, OR 1.09 (1.03–1.15, about 9% higher). Another pooled 21 randomized trials and 30,850 people. Exercise plus education cut low-back-pain episode risk RR 0.55 (0.41–0.74, about 45% lower), evidence quality medium. Exercise alone RR 0.65 (0.50–0.86, about 35% lower), evidence quality low to very low. Education alone RR 1.03 (0.83–1.27), back belts RR 1.01 (0.71–1.44) — neither showed an effect. The UK Health and Safety Executive (HSE) official guidance also states that training alone cannot guarantee safe lifting. Good technique does not replace handling aids and redesigning tasks and loads.

First vitamin D. A 2023 systematic review included 14 studies and 2734 fracture patients. The authors conclude that vitamin D alone has little effect on bone healing, union, or function after recovery. Studies that seemed to show benefit were generally lower quality. They state that counting only randomized trials, none showed a benefit from supplementing. One enrolled 100 long-bone fracture patients with low vitamin D: half took 100000 IU once within two weeks of injury; half took placebo. Two in each arm failed to unite — 4% together. Another enrolled 102 people aged 18 to 50 with thigh or lower-leg fractures in four dose or placebo groups. At 3 and 12 months, healing scores showed no difference. Another enrolled 32 postmenopausal women with wrist fractures. Versus no supplement, a high-dose group equivalent to about 1800 IU a day had lower compressive stiffness of bone. Next calcium. A 2014 systematic review says clinical studies of how vitamin D insufficiency or supplementing affects fracture healing are few and conclusions unclear. The two studies it found used calcium plus vitamin D together. They saw higher bone density or callus (new bone) at the fracture site; they did not look at healing speed. Finally bone broth. A Kaohsiung, Taiwan assay bought three bone-broth noodle soups, 450 to 550 g of broth per serving. Mean calcium per serving was 14 to 39 mg — 1.4% to 3.9% of a 1000 mg daily recommendation. The authors conclude one serving supplies under 5% of the daily recommendation. China’s recommendation is 800 mg a day at ages 18 to 49 and 1000 mg from 50 up; daily upper limit 2000 mg. The health industry standard that sets these figures also states that long-term excess calcium raises kidney-stone risk. A U.S. trial of 36282 postmenopausal women taking 1000 mg calcium plus 400 IU vitamin D daily, followed a mean 7 years: kidney-stone risk about 17% higher (HR 1.17, 95% CI 1.02–1.34, credible range). Hip fracture about 12% lower, not significant (HR 0.88, 0.72–1.08). It looks worthwhile because of the intuition “bone is made of calcium, so more calcium grows faster.”

First medicine. A 2026 Nature Medicine randomized trial enrolled 1298 ordinary people in the UK; the protocol was preregistered. Each got two prepared cases and had to judge the disease and whether to go to emergency or watch at home. Some used an AI chatbot — GPT-4o, Llama 3, or Command R+. Controls used their usual methods, mostly web search or the NHS site. Handed a full case directly, AI named a relevant disease 94.9% of the time and picked the right care setting 56.3% of the time. Ordinary people using the same AI named a relevant disease under 34.5% of the time and picked the right setting under 44.2%. Controls recognized a relevant disease 1.76 times as often as AI users (95% CI 1.45–2.13, credible range). Recognizing red-flag serious illness was 1.57 times (1.28–1.92). Judging where to go was similar on both sides. The authors reviewed chat logs; problems were two: users did not tell all symptoms, and when AI named the right disease users did not take it. Next law. A U.S. research team in 2024 asked public models including ChatGPT 4 about U.S. federal case law. ChatGPT 4 had fabricated content in 58% of answers; Llama 2 up to 88%. Fabrication means content that does not match legal fact; the paper calls it “hallucination.” Models cannot tell which of their own lines are made up. When the user’s question carries a wrong legal premise, they often accept it wholesale. The same team in 2025 tested professional legal AI tools for lawyers. Tools that claim not to fabricate still had fabrication in 17% to 33% of answers. It looks worthwhile because AI is always on, free, and speaks with a doctor’s or lawyer’s certainty.

First how long people live. WHO’s Global Health Observatory computed a 2023 life table for China. A life table uses that year’s death rates by age to project how long a group lives. By that table, probability of a 30-year-old living to 60 is 92.0% (men 89.2%, women 94.9%). To 65: 87.4% (men 83.0%, women 92.0%). To 80: 55.6% (men 45.8%, women 66.1%). These three sets are ratios of numbers still alive at each age in the table. People who reach 65 have on average another 17.5 years. Next how people estimate. A 2019 Journal of Finance study surveyed people on how long they expect to live. Young people underestimate survival; older people overestimate. An earlier working-paper example: a 28-year-old man’s true 5-year survival was 99.4%; he estimated 92.8%. The authors put self-estimated probabilities into a standard life-cycle saving-and-spending model. Versus using actual death rates, young people save 26% less and retirees draw down savings 27% more slowly. People whose life-and-death estimates in the survey were more skewed also saved more skewed, after subtracting risk preference, cognition, and household conditions. The authors think one reason is overweighting rare, memorable ways of dying.

First osteoarthritis. A 2002 trial split 180 people with knee osteoarthritis into three groups: arthroscopic clean-out, arthroscopic lavage, and sham surgery. Sham only cut a few skin holes; the scope was not inserted. Patients and assessors did not know who got which. At every follow-up over two years, neither real surgery beat sham on pain or function. Knee-pain score is 0 to 100; higher hurts more. At one year: sham 48.9, lavage 54.8, clean-out 51.7. A 2008 trial enrolled 178 people with moderate-to-severe knee osteoarthritis. One group had arthroscopic clean-out plus physiotherapy and drugs. The other only physiotherapy and drugs. Symptom total is 0 to 2400; higher is worse. At two years the groups were 874 versus 897, difference 23 (95% CI −208 to 161, credible range) — no difference. Next meniscus. A 2013 Finnish trial enrolled 146 people aged 35 to 65 with symptoms of a degenerative medial meniscus tear and no osteoarthritis. One group had partial meniscectomy; one had sham surgery. At 12 months three main scores did not differ. For example post-exercise knee pain (0 to 10): groups improved 3.1 and 3.3. A 2015 meta-analysis pooled 9 trials. People who had surgery had 3 to 5 points less pain at 3 to 6 months (0 to 100). By 1 to 2 years there was no difference, and function did not improve. Next harms. Symptomatic deep-vein thrombosis in the leg: about 4.13 per 1000 operations (1.78–9.60). Also pulmonary embolism, infection, and death. A 2017 BMJ international expert panel therefore strongly opposed arthroscopy for almost all people with degenerative knees. The panel states recovery takes 2 to 6 weeks after surgery. Finally what to do. A 2016 Norwegian trial enrolled 140 middle-aged people with degenerative meniscus tears. One group did 12 weeks of supervised exercise; one had partial meniscectomy. At two years knee scores differed by only 0.9 (−4.3 to 6.1) — no difference. At 3 months the exercise group had stronger thigh muscle. The 2015 Cochrane systematic review pooled 44 trials and 3537 people. Exercise cut knee pain by the equivalent of 12 points (0 to 100, credible range 10 to 15), evidence quality high. Two to six months after stopping, still 6 points less. A 2013 U.S. trial enrolled 454 people aged 55 and above, overweight or obese, with knee osteoarthritis. Diet plus exercise lost a mean 10.6 kg (11.4% of body weight). At 18 months their pain score (0 to 20) was 3.6 versus 4.7 for exercise alone. It looks worthwhile because “cleaning out worn and torn bits” sounds on-target, and many people do feel better after surgery.

First how large the effect is. A 2013 meta-analysis pooled 19 trials and 1683 people. Each trial was randomized versus placebo. Sleep problems were not caused by another disease; 14 of the trials studied insomnia. Melatonin shortened time to sleep onset by a mean 7.06 minutes (95% CI 4.37–9.75, credible range). Total sleep time rose 8.25 minutes (1.74–14.75). Sleep-quality improvement was also small (SMD 0.22, 0.12–0.32). The authors themselves note this shortens onset less than other sleep drugs. Next guidelines. The American Academy of Sleep Medicine’s 2017 chronic-insomnia medication guideline advises clinicians not to use melatonin for hard-to-fall-asleep or restless sleep; recommendation strength “weak.” Basis: 3 trials, all using 2 mg sustained-release melatonin, all subjects aged 55 and above. Evidence quality “very low.” The only poolable outcome was sleep quality (SMD 0.21, −0.36 to 0.77), short of clinical meaning. Melatonin health foods in China list the function as “helps improve sleep.” Precautions say “this product cannot replace drugs”; unsuitable groups are children and adolescents, pregnant women, and nursing mothers. A 2026 Beijing market-regulation science piece also says it is “by no means a ‘wonder drug’.” Jet lag is separate. A Cochrane systematic review included 10 trials; 9 of them shared one finding. After flying across 5 or more time zones, melatonin at local bedtime eases jet lag. Daily 0.5 to 5 mg works about the same. It looks worthwhile because it is a hormone the body already makes, sounds mild, and needs no prescription.

No study specifically compared IV versus no IV for a cold, so this item rests on official documents and monitoring data. First colds. Guangdong’s health and family-planning commission issued a 2017 document requiring primary facilities to cut unnecessary IVs. It states drugs should follow WHO’s principle: “oral if possible; intramuscular if injection is needed rather than intravenous.” Only three situations use an IV: swallowing difficulty; severe absorption problems such as vomiting or severe diarrhoea; and critical, fast-moving illness that needs high drug levels in the body quickly. An attached list of common outpatient and emergency conditions that in principle need no IV starts in internal medicine with “upper respiratory infection: ordinary cold, viral pharyngitis.” Next “flushing vessels.” The National Health Commission held a briefing on 6 August 2025. A chief physician from Dongfang Hospital, Beijing University of Chinese Medicine, clarified the rumour. The quote: “The claim that seasonal IV drips can clear vessels and prevent cerebral infarction is wrong.” Drips cannot “wash away” or “clear” atherosclerotic plaque. Blind drips may add load on heart or kidneys and even trigger heart failure. Finally risk. The National Adverse Drug Reaction Monitoring Annual Report (2024) breaks reports down by route. Of reported reactions, injection routes were 57.2% and oral 33.1%. Among injections, intravenous was 91.0%. That is the share of reports, not the chance of harm per drip. It looks worthwhile because the bottle is visible, you lie there until it finishes, and it feels as if the drug arrives faster and fuller.