59 findings · 16 substances · 22 well-evidenced
Fixing a real shortage and making up for lost sleep account for the energy gains on record. For ordinary tiredness in people short of nothing, the evidence is thin, contested or was never collected.
Deficiency is judged on four markers, not one: the blood level, the share bound for delivery into cells, and two that build up when cells cannot use what is there.
This is a snapshot of one Irish and Northern Irish cohort recruited between 2008 and 2012, not a picture of what happens over time. Anyone already taking a B12 supplement was excluded, which sharpens the dietary signal but leaves out most real supplement users.
Over 4.3 years against placebo, metformin cut B12 by 19 percent and left 7.2 percentage points more people deficient. Because people were randomised, it is the drug doing it.
Everyone in the trial was on insulin for type 2 diabetes, so the absolute numbers may not carry to metformin alone. Deficiency was called on the blood level alone, not on markers of whether cells were short. The homocysteine rise that would prove a functional cost fell short of significance.
At five years, low or borderline B12 was twice as common on metformin as on placebo, 19.1 percent against 9.5, and each extra year raised the odds 13 percent.
B12 was measured only at years 5 and 13, so nothing is known about the years between. Frank deficiency at 13 years, 7.4 percent against 5.4, did not reach significance. The neuropathy link is an observation within the group, not a demonstration that the low B12 caused it.
Oral treatment works at 1 to 2 milligrams a day, far above an ordinary supplement. At that dose, over 90 percent of both the tablet and injection groups had normal levels by week eight.
These trials counted blood levels, not whether anyone felt better. The Cochrane review behind the comparison had three eligible trials and 153 people, and graded its own certainty low. By one year the oral group had slipped slightly behind the injected one on the stricter analysis.
Pooling 17 studies, vegan adults had significantly less B12 in the blood than omnivores and significantly more homocysteine, which builds up when cells cannot get B12.
The two most specific markers, methylmalonic acid and holotranscobalamin, pointed the same way but missed significance at p=0.06 and p=0.09, so the case rests on homocysteine, which folate and kidney function also move. The comparisons are observational, so other diet-linked differences remain live.
In people with high blood pressure, 200 to 300 mg of caffeine raised it for three hours or more. Over about 2 weeks, coffee raised it no more than going without caffeine.
The hour-by-hour rise comes from single doses given in short test sessions, which may not match a regular drinker's morning cup. Only three trials, of about 2 weeks each, looked past the first day, too short to catch a change that builds more slowly. An older review found the rise strongest in people who already have high blood pressure, so it may be smaller in others.
Across 13 studies of 314,827 adults followed for years, coffee drinkers got high blood pressure no more often than others. A 10-year study using 24-hour readings agreed.
Coffee habits were self-reported and chosen, not assigned. People worried about their blood pressure may cut back, which would hide a rise among heavier drinkers. An earlier pooling of six studies found 9% higher risk at 1 to 3 cups a day and no rise above 3. American studies found slightly lower risk and European ones did not, so the overall figure averages a real split.
Pooled across eight placebo-controlled trials, caffeine left blood pressure about 2 mmHg higher, more so in men, at higher doses and in trials past nine weeks.
The trials gave caffeine itself, not coffee, at doses and for lengths that varied from trial to trial. Eight trials is a small base to split by sex, dose and length, so those three patterns are leads rather than settled findings. The clearest dose effect, above 400 mg a day, was in diastolic pressure.
Pooling 45 studies, caffeine restored reaction time, attention and driving in people cut to six hours of sleep or kept awake a full day, and bigger doses brought bigger gains.
Everyone in these 45 studies was short on sleep, at six hours or under or awake for a full day, so nothing here says what caffeine does for someone who slept well. The benefit kept rising with dose without the review ever finding the point where it stops rising, so the top of that range is unmapped.
European regulators advise under 200 mg of caffeine a day during pregnancy, against 400 mg for other healthy adults, and no more than 200 mg in a single serving.
This is a precautionary ceiling drawn from a broad safety review, not a figure any trial in pregnant women produced. It describes a population rather than any particular pregnancy.
Two agencies reached 400 mg a day, and 200 mg in one sitting, separately, from a review of more than 300 human and animal studies. It is a population figure, not a personal one.
This is a threshold for a population of healthy, non-pregnant adults rather than a personal limit; the same 400mg lands differently on people who clear caffeine at different speeds. It is regulatory guidance rather than the result of a single trial. The 200mg single-serving figure rests on thinner data than the daily one.
Pooling 11 trials of caffeine-theanine blends, the gain in alertness over the first two hours tracked how much caffeine was in the mix more closely than how much theanine.
The 11 trials used a spread of doses and designs rather than one fixed ratio, so this is a pattern across mixtures rather than a test of any single blend. The authors call theanine's contribution comparatively modest rather than absent, which leaves room for a larger pool to move it. And every trial pooled here tested combinations, so none of it measures theanine by itself.
A daily 100 mg of caffeine, about one cup of coffee, is the lowest intake shown to do it. Symptoms start 12 to 24 hours after the last dose and can run nine days.
Those timings are population ranges, and plenty of people stop and feel nothing at all. The 100mg figure is the lowest daily intake shown to produce withdrawal, not the usual one. This is a critical review of 57 experiments and 9 surveys rather than a pooled statistical analysis, though nothing since has replaced it.
Pooling 15 studies and 186 people who began 3.6 percent lighter than normal, aerobic performance fell 2.4 percent. How dry they were did not change how much they lost.
An average 3.6 percent of body mass is more dehydration than most recreational exercisers ever reach, and the performance tests averaged 22 minutes. The dehydration was deliberately induced, usually by heat or fluid restriction, which brings its own fatigue on top of the water deficit. The confidence interval on the headline runs from 0.8 to 4.0 percent. Nothing here compares an electrolyte drink with water.
In 150 young women who were not anaemic, ferritin finished at 44.8 against 43.8 micrograms per litre. Stomach side effects were 1.56 times as common on the daily schedule.
The trial's one favourable repletion number is a secondary outcome at p=0.049 resting on roughly 2 women against 8, which is too few to build on. It ran at a single centre in 150 young Swiss women, none of them anaemic, so it does not establish the same schedule in men, in older adults, or in anyone with iron-deficiency anaemia.
Seven of 38 blood-donor trials measured deferral for a low haemoglobin, and iron cut it by 61%. Haemoglobin, ferritin and a rise of about 69% in stomach complaints were all rated very uncertain.
Doses across the pooled trials ran from 7 to 1,800 milligrams of iron a day, schedules that are not clinically interchangeable. 27 of the 38 studies left out first-time donors, who are the people least likely to come back after being turned away and therefore the ones the intervention most needs to reach. The review was produced by a national blood service, which has a direct operational interest in donor retention.
Against bone marrow iron, a ferritin under 30 micrograms per litre caught 79% of the people who were deficient and wrongly flagged 2% of those who were not.
The 30 microgram figure rests on 9 studies and 512 people, all graded low or very low certainty. The studies underneath it proposed thresholds anywhere from 12 to 200 micrograms per litre. And ferritin was far worse at spotting iron overload than deficiency.
In young women low on iron, 40 mg delivered 6.7 mg absorbed and 240 mg only 18.1 mg. Doses of 60 mg and up raise a hormone that shuts absorption down for a day.
Fifty-four young Swiss women, split across three sub-studies, all with low iron stores and none anaemic. What was measured is how much iron crossed the gut over a day or two, not haemoglobin or stores rebuilt over months. The paper argues the hepcidin timing supports alternate-day dosing but never tested such a schedule.
In UK Biobank men with two copies of the C282Y variant, 20.3% had liver disease against 8.3%, and 33.1% died against 25.4%. One copy carried no excess of anything.
People carrying only one copy, and those with the milder H63D variant, showed no excess of anything — and they far outnumber the two-copy group, which is roughly 1 in 150 to 200 people of northern European ancestry. UK Biobank volunteers are healthier than the general population, so these percentages are not population rates. The one-in-ten lifetime figure is for men; the figure for women is substantially lower.
Both halves come from the same 67 trials in 8,506 menstruating women. Haemoglobin rose and fatigue eased; loose stools and constipation each roughly doubled.
Only 10 of the 67 trials were judged at low risk of bias. The harm figures rest on far fewer women than the benefit ones — 521 to 604 against 6,861 for haemoglobin — so they are much less precise. The review's composite measure of any side effect at all was not significant, even though the specific ones were.
In 440 adults with iron deficiency anaemia, haemoglobin at two weeks rose 2.00 grams per decilitre with vitamin C and 1.84 without. Ferritin at eight weeks matched too.
The iron dose was 100 milligrams every eight hours, high enough that absorption is already saturating and there is little room for an enhancer to help. Everyone in the trial was frankly anaemic, in whom absorption is already turned up. And this tested a vitamin C tablet taken with an iron tablet, not the vitamin C in a meal acting on the iron in food.
In men over 65 treated for a year, sexual activity, desire and erectile function all improved. Fatigue and walking distance did not, and pooled trials found nothing on energy.
About 65 men were screened for every one enrolled, so these are not the men who walk into a clinic. One year of treatment is short against how long testosterone is actually taken. In the pooled analysis the sexual-desire interval ran from 0.01 to 0.34, close enough to zero to be worth naming.
Eating five or more portions of B12-fortified food a week went with better markers across 3,299 older adults. In the 15 percent whose stomach lining had thinned, it was not enough.
This is an association measured at a single point in time, so it cannot show that the fortified food did the lifting. The thinned stomach lining was inferred from a blood ratio rather than confirmed by biopsy, so the group is drawn approximately.
About 6 percent of over-60s are outright deficient and roughly 20 percent sit in the marginal band. The usual cause at that age is not a meat-free diet but food-bound malabsorption.
The survey figures were reviewed in 2009 and predate current fortification and supplement habits. The 148 pmol/L line between deficient and marginal is itself contested, and moving it moves both percentages. The review was written to argue for fortifying flour, a stated policy position.
In one 12-year study of 155,594 US women, drinkers of cola, sugared or diet, were more likely to develop high blood pressure. Coffee drinkers were not.
Something else in cola, or about the people who drink it, may carry the risk, and the authors could not say what. Caffeine intake was not flat either: women in the middle fifth had 12 to 13% higher risk than the lowest. Both intake and diagnosis were reported by the women themselves.
In one trial, young men drank a litre of black coffee with 212 mg of caffeine and passed no more urine over four hours than on water. Tea and cola, with less caffeine, did no worse.
With 17 men on coffee, the trial was sized to catch a difference of 220 ml of urine or more, so a smaller loss could have gone unseen. Larger doses were not tested, and the researchers note a measurable effect on urine is likely once a single dose passes about 300 mg of caffeine. The men were fully hydrated and had fasted overnight. The litre went down in half an hour, not as cups spread across a day.
Cordycepin, the compound the marketing rests on, is taken apart by an enzyme your body already makes, leaving it a short half-life and low bioavailability.
The pharmacology paper is a review and reports no human half-life figure. Its interest is in engineering cordycepin into something usable, with an enzyme inhibitor, a redesigned molecule or a new delivery system, none of which exists in a capsule. And it never asks the separate question of whether the amounts in a supplement would matter even if they did survive.
Supplements led on raw cordycepin at 81.4 mg per 100 g, but run through a simulated gut the fruiting bodies released the most, about 26 mg per 100 g.
Two commercial supplements were tested, so this describes two products rather than a market. The digestion was simulated in a laboratory model, which estimates what could be released into the gut, not what reaches the bloodstream or any tissue. No health outcome was measured anywhere in the study.
Cordycepin and mannitol turned up in all ten samples, the five authentic and the five substitutes alike. Mannitol alone ranged fifty-sevenfold across them.
Ten samples in total, five per group, which is small for a statement about a market. Two of the authors work for a company selling authentication testing, which is a commercial interest in the standard markers failing. The paper set out to propose better markers, so that failure is its starting premise rather than a measured rate.
DNA testing of supplements labelled Ophiocordyceps sinensis, the costly wild caterpillar fungus, turned up the cheaper Ophiocordyceps militaris instead.
Both are method papers: they show substitution happens without reporting how often it happens across a defined slice of the market. The products were bought in Asia, so the rate may not carry over to Western shelves. And DNA says which species is present, not how much of the product it accounts for.
The five tested 321 people aged 16 to 35 at doses spanning twelvefold, 1 to 12 grams. Some found gains in aerobic performance; across the five the results were inconsistent.
The review appraised each study for bias but never pooled them, so there is no combined estimate here, only five separate results. It covers Cordyceps militaris alone, leaving out both the wild caterpillar fungus and the fermented mycelium products that fill much of the retail shelf.
Paecilomyces hepiali, one of the commonest label names, is a fungus isolated from wild Chinese cordyceps rather than that organism itself.
This is a review of the Chinese industry rather than Western retail, and it describes those names rather than counting how often they appear. One author works for a cordyceps manufacturer, though the paper argues against loose labelling rather than for it. The authors say two of the fungi they discuss need further study before their place is settled.
Pooled trials in people without anaemia found gains of 0.34 to 0.53 standard deviations in fatigue, reasoning and memory. Leave out those who were iron deficient and the gains go too.
The subgroup finding that carries this claim is reported without a formal test of interaction, so it is suggestive rather than demonstrated. Six of the eighteen studies had no control group, and their effect sizes run two to three times larger, which is what an uncontrolled design predicts. Only 204 of the 1,408 participants were children or adolescents.
Absorption runs 14 to 18% on a mixed diet and 5 to 12% without meat, but both figures describe people whose stores are already empty. As those stores fill, the percentages drop.
These are modelled ranges built from intake data and isotope studies, not measurements taken in one population, which is why they are given as spans rather than single numbers. The review is a narrative one and lists the nature of the meat factor, and the roles of vitamin A and carotenoids, as still unresolved.
US labels call accidental iron overdose a leading cause of fatal poisoning in under-sixes. 195,780 children's iron-vitamin ingestions killed nobody.
Poison-centre and poisons-service data are voluntary referrals rather than a census, so a count of zero deaths describes the cases that reached them. The UK cohort counts hospital referrals, which would miss a child who died before or too soon after arriving. And the US figures stop in 1998, straddling the 1997 packaging rule without separating out its effect.
The best and the worst single meal for iron absorption differ 4.4-fold. Compared across a diet eaten over two weeks, that gap falls to two-fold.
This is a narrative review, so which studies went into it was the authors' judgement, and they work at the laboratory that produced much of the underlying isotope work. The two-week comparison rests on a single cited study. And single-meal effects are real, so this deflates the advice rather than cancelling it.
Among unwell people with no iron left in their marrow, ferritin averaged 82 micrograms per litre, above the 70 seen in healthy people who still had some.
Every study underneath these means was observational. The 70 micrograms per litre comparator rests on three studies and 151 people. The inflammation correction is a statistical adjustment to survey data rather than a check against bone marrow, and the two proteins it needs are not measured alongside an ordinary ferritin test.
The number printed largest on the bottle is the weight of the whole compound. A standard 325 milligram ferrous sulfate tablet delivers 65 milligrams of elemental iron.
This is one manufacturer's label. Other iron salts — fumarate, gluconate, bisglycinate — deliver different amounts of elemental iron per tablet, so the ratio here is not a conversion factor for every product. The label itself carries three disagreeing revision dates, so treat it as a record of what a shopper is told rather than of when they were told it.
Of 1,839 single-substance iron overdoses referred to the UK poisons service, 28% were children under six and 37% were aged 13 to 20, 91% of them girls.
The authors read the older peak as intentional rather than accidental, which is an interpretation of the age and sex pattern rather than something the referrals recorded. These are hospital referrals to a poisons service, not all iron overdoses. And how much iron someone said they had taken predicted neither their measured blood level nor their symptoms, so the exposure data here are unreliable.
The recommendation is 8 milligrams a day for men, 18 for women who menstruate and 27 throughout pregnancy. Median intake runs about 12 for women and 16 to 18 for men.
A median is not a person: half of women eat more than 12 milligrams and half less, and falling below a recommendation is not the same as being deficient. These intake figures were compiled for a 2001 report and have not been revised since. The absorption assumptions behind the recommendations predate what is now known about how the body regulates uptake.
The adult upper limit came from one 97-person trial of stomach side effects, divided by an uncertainty factor of 1.5. One ordinary ferrous sulfate tablet supplies 65 milligrams.
People with hereditary haemochromatosis were deliberately left out of the derivation as exceptionally sensitive, so this figure does not protect them. The same 45 milligrams was applied to pregnancy only because data on lower doses in pregnant women did not exist. The chapter was published in 2001 and has not been revised since.
In a national sample of 8,021 adults, 14% had ferritin below 30 nanograms per millilitre. Of the women who were deficient, 22 to 35% took any iron at all.
A single blood draw cannot say who felt unwell or who would have benefited from treatment. The headline figure moves with the cutoff chosen, and below 30 is one of several in use. No inflammation adjustment was applied, which by other work would push the estimate higher. The authors say what functional deficiency means in an otherwise well adult is still unresolved.
The evidence disagrees with itself, or the finding carries a caution. Either way the marker shows one star, however strong the individual sources are.
The sixteen trials pooled 6,276 people who were not deficient and found nothing on memory, attention or depression. Fatigue was meant to be the third outcome, and there was nothing to pool.
Everyone studied already had normal B12, so this says nothing about people who are actually short of it. Several of the pooled trials gave B12 inside a B complex alongside folate, so the B12-only signal rests on fewer studies. The fatigue gap is an absence of trials, not a pooled null.
Eighteen pooled trials found a moderate drop in self-reported tiredness and no change in anything objective. The two trials that raised iron stores furthest found nothing at all.
The pooled fatigue result rests on 4 of the 18 trials, on self-reported scales, and iron is hard to blind because it darkens stool. On the other side, both null trials ran in blood donors, a healthy group whose average fatigue was about 4 out of 10, leaving less room to improve. One of them lasted four weeks, and its own authors note the donation had not made these women noticeably tired.
In people low on iron but not anaemic, iron refills the stores. Whether that eases tiredness is open: pooled trials found people reporting less of it, while two placebo-controlled trials, one refilling stores more fully, found nothing.
Over 12 weeks in 120 overweight adults aged 40 to 75, stress scores fell about as far on placebo as on ashwagandha. Fatigue fell further on ashwagandha than on placebo.
Placebo groups in earlier ashwagandha stress trials barely moved, falling around 5.5%, so a 36% placebo fall may be what hid the difference here rather than a weaker herb. Cortisol was not measured, so this says nothing about the hormone. The fatigue result was one secondary outcome among many measured, which makes a chance finding more likely. The trial was paid for by the extract's maker.
Whether ashwagandha's edge on stress survives a strong placebo response is open. Pooled trials found it beat placebo, but in older, overweight adults the placebo group improved by a third and matched it.
In 95 people with disabling tiredness and normal B12 levels, eight weeks of 1000 micrograms a day and a placebo each improved fatigue by about eight points.
Both arms improved by roughly eight points, a large placebo response that leaves little room to detect a small real benefit. The bar for normal B12 was set at 150 pmol/L, low enough that some participants may have been functionally short despite passing it.
Cyanocobalamin at 1 to 2 milligrams a day cleared deficiency, and a separate trial of methylcobalamin saw the same markers fall. No trial gave the two forms to the same people.
None of these trials set out to compare the forms, so this is an absence of evidence that one beats the other, not evidence they are the same. They enrolled different patients on different schedules, which is what a real comparison would have to hold constant.
In 77,118 adults, men taking over 55 micrograms a day from a standalone supplement for a decade had roughly twice the lung cancer risk of non-users. Women showed nothing.
It appeared only for standalone high-dose supplements, not for the same vitamins inside a multivitamin, which is hard to explain if the vitamin is doing it. Intake was reported once, at the start. The review clearing B12 includes authors with declared ties to manufacturers.
Whether high-dose B12 supplements raise lung cancer risk is unresolved. One cohort saw about double the risk in men, none in women, a split nobody has explained, and the risk ran higher in smokers, so something besides B12 may explain it.
About 5 percent of a large oral dose is absorbed, and no tolerable upper intake has been set, for the general population or for pregnancy. Harm has not been demonstrated.
Both figures are background citations inside a review about pregnancy trials rather than results its authors produced, so they carry the weight of a textbook footnote and not of a measurement. The review dates from 2012 and predates the later cohort reports on high blood B12 and cancer.
In pernicious anaemia the body stops making the protein that carries B12 across the gut wall. Twenty-six patients swallowed 1000 micrograms a day, and after a month 23 were no longer deficient.
Twenty-six people, one hospital, and no group given injections to compare against. Three of them did not respond. The dose works by seeping across the gut wall rather than through the missing transporter, so a run of missed days costs more than it would with a monthly injection.
Whether a swallowed B12 dose is enough when the gut cannot absorb it normally is still open. The support is 26 patients with no comparison group and three small trials rated low certainty: promising, not yet properly tested.
Across 25 observational studies, people on proton pump inhibitors had 42 percent higher odds of B12 deficiency, but most of the studies found no difference in blood B12.
Every included study was observational, and people on long-term acid blockers differ from those who are not in ways that also affect B12. Heterogeneity was high, with just over half the variation between studies not attributable to chance. The review was not registered in advance.
In one trial of 11 adults taking caffeine daily, its blood-pressure rise had faded to near zero by day 20. The nervousness and poor sleep had not.
Eleven people give only a rough picture of when the rise fades and how completely. One moderate dose was tested, about 210 mg for a 70 kg adult, not the range people drink.
In one small trial, daily caffeine's blood-pressure rise faded to near zero within three weeks; pooled trials found about 2 mmHg that was larger in longer trials. Whether the body adapts fully or only partly is unsettled.
In three trials of 135 people in all, oxygen uptake, ventilation and anaerobic power sat where placebo left them. Everyone studied was young and already trained.
Two of the three tested multi-ingredient products, one of them eight ingredients deep, so none of this isolates cordyceps from what it was mixed with. Every participant was young and already trained, which is where there is least room left to improve. The shortest of the three ran fifteen days.
Whether cordyceps improves endurance is unsettled. A pooled analysis reports gains that only just clear chance, while the three best-powered single trials found none, which says more about how thin this research is than about who is right.
Lovastatin measured 53.6 mg per 100 g of dry weight in the two retail Cordyceps militaris products analysed, and nobody buying them was told it was there.
Two products were tested, so this is a finding about two capsules on a shelf, not a rate across the market. The study measured content and stopped there: nobody swallowed anything, and no effect on cholesterol or on anything else was looked for.
Across human trials of two to sixteen weeks people held out longer before stopping, and the longer they supplemented the bigger that gain. Aerobic fitness did not follow.
This is a narrative review with no stated search protocol and no bias appraisal, so it carries forward the weaknesses of the trials it summarises rather than weighing them. Almost everyone studied was young and active. The authors say plainly that older and sedentary people are barely studied.
Twenty-six triathletes took 2.6 g of sodium in capsules or a dummy through a half-ironman. The salt group finished sooner. Strength, jump height and sweating were unchanged.
The finish-time result is a single p-value of 0.04 among eight reported outcomes, the shape of finding that most often fails to repeat. The abstract never says how many minutes faster, so the size of the advantage is not on the record anywhere. The trial's registry entry declared 40 participants against the 26 analysed, unexplained. Four to six hours of racing by trained triathletes says nothing about an hour in a gym.
One randomised trial found salt capsules got triathletes to the finish sooner, on a single borderline result in 26 people. Whether that holds up is open, because nobody has repeated it.
The panel behind the US iron upper limit found no evidence of clinically significant harm from the pairing. Pooled trials in under-threes saw zinc move iron markers the wrong way in a subgroup.
Neither source was designed to test the pairing: one is a set of marker outcomes graded very low to moderate certainty, the other an absence noted while setting an intake limit. The child trials used substantial doses of both minerals in populations where deficiency is common, not the amounts in an ordinary multivitamin. And the panel's finding was recorded in 2001 and has not been revisited.
A US panel found no evidence of clinically significant harm from iron and zinc interacting; pooled trials in under-threes saw zinc move iron markers the wrong way. Neither set out to test the pairing, so whether to space them is unknown.
The trials recruited people with multiple sclerosis, post-COVID fatigue or cancer, or with levels below the deficient line. None enrolled someone simply tired with a normal level.
The four-week trial scored fatigue on a self-reported questionnaire, a measure that moves easily on expectation, in people averaging 29 years old. The multiple sclerosis pooling covered five trials and 345 people, and its benefit runs from small down to almost nothing. The palliative trial lost 94 of its 244 participants, mostly to death from cancer.
In one trial of 27, nine per group, slow-release caffeine cut daytime sleepiness but tended to worsen sleep; melatonin helped sleep but not daytime sleepiness.
The trial reported which way each measure moved, not by how much. Nine people took each treatment, and three published papers all describe the same 27. The caffeine was a slow-release tablet, not coffee.
Search runs entirely in your browser. Nothing you type is sent anywhere.