Water
Most adults walk around mildly under-hydrated, and the cost is bigger than you'd think — sitting at the high end of the normal blood-sodium range tracks with faster biological aging and a measurable rise in cardiovascular and dementia risk decades later. The fixes are unglamorous: drink steadily through the day, filter out the industrial contaminants, leave the calcium and magnesium in.
For a long time, hydration science was binary: you were either fine or clinically dehydrated, and anything in between didn't matter. A 2023 analysis of the Atherosclerosis Risk in Communities (ARIC) cohort — 15,752 US adults followed for 25 years — put a stake through that view by showing that serum sodium levels at the high end of the "normal" reference range predict accelerated biological aging, chronic disease, and earlier death — with habitual low fluid intake the most likely upstream driver, though the cohort cannot prove that.[1] A 2026 analysis of the same cohort plus 395,802 members of an Israeli health service put a number on the dementia limb for the first time: risk starts climbing around 141 mmol/L and runs 1.5- to 3-fold higher at the top of the range and above it.[2] Add the modern reality of per- and polyfluoroalkyl substance (PFAS) and microplastic contamination of both municipal and bottled water, and "drink water" stops being a one-line piece of advice.
Hydration as a longevity intervention
Evidence: Moderate — one large prospective cohort on serum sodium, corroborated but not independently replicated, plus an independent cohort using a different hydration measure, with a plausible mechanism, but no trial has shown that drinking more slows aging or extends life.
The cleanest evidence linking lifelong hydration habits to lifespan comes from cohorts that use serum sodium as a stable physiological proxy for chronic fluid intake. Lower water intake → less total body water → higher serum sodium. ARIC enrolled 15,752 middle-aged adults and followed them for 25 years, scoring an algorithmic "biological age" from a panel of 15 organ-system biomarkers, nine of which carried the calculation. That biological-age analysis ran in a healthier subset of 6,956 taking neither blood-pressure nor cholesterol medication, and the chronic-disease analysis in the 5,168 participants who survived to the final visit.[3]
| Serum sodium (mmol/L) | What it means |
|---|---|
| 138–142 | Closest to the trough. The lowest biological age sits at about 139 mmol/L, and the curve is U-shaped rather than flat — biological age rises below roughly 139 as well as above it.[4] Dementia risk begins climbing around 141 mmol/L, inside this band.[5] |
| >142 | About 10–15% higher odds of being biologically older than chronological age, rising to roughly 50% only above 144 mmol/L. Roughly 39% higher risk of a composite of heart failure, dementia, chronic lung disease and stroke, measured against the 138–140 mmol/L group and in the 5,168 participants who survived to the 25-year visit; adding type 2 diabetes, peripheral artery disease and atrial fibrillation was a sensitivity analysis. |
| >144 | About 21% higher premature all-cause mortality than the 137–142 mmol/L reference band — 34.5% of that group died during follow-up against 26.2% of the reference group — though the estimate only just clears no effect. |
Critically, the whole of 135–145 mmol/L is reported as normal by a standard laboratory. The risk gradient sits almost entirely inside that window. The same signal shows up in an Israeli cohort of 407,187 adults healthy at baseline: sodium of 140–142 mmol/L carried about 13% higher risk of later hypertension, rising to about 29% above 143 mmol/L, with heart-failure risk about 20% higher in that top band — and among the health service's current healthy members, roughly six in ten sit at 140 mmol/L or above.[6] Note that the two cohorts put the inflection in different places — ARIC's optimum runs to 142 mmol/L, while the Israeli data show risk climbing from 140 — so the low 140s are best read as a grey zone rather than a safe plateau. And read the second cohort as corroboration rather than independent replication: it shares its senior authors with the ARIC analysis. No genuinely independent cohort has yet reproduced the serum-sodium result itself. The closest independent support uses a different exposure: among 63,488 Japanese adults followed for a median of 19.4 years, the lowest fifth of estimated daily water turnover carried about 26% higher all-cause mortality in women and 18% in men, on an L-shaped curve that argues against being under-hydrated rather than for drinking ever more.[7] The 2026 dementia analysis cited above comes from the same group and the same two cohorts, so the same caveat applies to it; a large independent Dutch population cohort that tested serum sodium against new dementia directly, 8,028 adults followed for about 11 years, found no association — though it was designed around low sodium and modelled the rest of the range as a straight line, so it cannot rule out a threshold at the high end.[8]
The mechanism: vasopressin / copeptin
When body water is chronically low, plasma osmolality rises, and the hypothalamus drives the posterior pituitary to release arginine vasopressin (AVP). Short term, AVP saves you by concentrating urine. Long term, sustained AVP may be costly. It keeps the kidney concentrating urine continuously, and copeptin — AVP's stable surrogate biomarker — tracks with later heart failure: about 63% higher risk in the top quarter of the copeptin distribution than the bottom, among 5,297 older Swedish adults followed for a median of 11 years. Per standard-deviation step up in copeptin, adjusting for kidney function trims the association only modestly, while adjusting for a heart hormone that tracks fluid volume removes about half of it, which suggests much of the copeptin signal is shared with ordinary fluid-volume physiology.[9] Copeptin also travels with arterial stiffness and coronary calcium in general-population imaging, though only cross-sectionally.[10] None of that is causal evidence: the proponents' own review puts it no more strongly than that persistent activation of these adaptive responses may be detrimental, with causation still under test in trials.[11] And when inflammatory and oxidative-stress markers were measured directly against this exposure, they accounted for about 1% of the sodium-to-mortality association — though in that analysis the excess mortality sat with low-normal, not high-normal, sodium.[12]
One caveat belongs here rather than at the end. Serum sodium in healthy people is defended tightly by vasopressin and the kidney, so it is a proxy for habitual fluid intake, not a readout of it. A published critique in the same journal presses the point further: blood was drawn after an overnight fast, so a high morning sodium partly indexes how well the kidney concentrates urine overnight, and six of the fifteen biomarkers behind the biological-age score are themselves readouts of kidney function — meaning poorer kidney function can push both sides of the association at once.[13] And the step from "high-normal sodium predicts disease" to "drinking more water prevents it" is an inference, not a finding. Three randomised trials have now tested that step against an organ outcome. The two in adults with kidney disease were null (see below). The third encouraged 175 healthy children, median age 8.6 years, to drink more for a year and found office blood pressure about 2 mmHg lower than in controls — small, unblinded, and attributed by its own authors to more efficient kidney sodium excretion rather than to vasopressin.[14] Read the table as a corroborated but not independently replicated association with a plausible mechanism, not as a demonstrated intervention.
What augmented water intake actually does
Evidence: Moderate for preventing repeat stones in people who already form them, and on a single trial for recurrent urinary tract infections; Weak for weight and everything else.
A 2024 JAMA Network Open systematic review pulled together every randomized trial that deliberately changed how much water people drank — 18 trials in all.[15] Its honest summary is narrower than the popular version: only weight loss and kidney stones had more than one trial pointing the same way — the review's own tables also record statistically significant single-trial results for recurrent urinary tract infections, fasting glucose and blood pressure. Ten of the 18 trials reported at least one positive result; eight reported negative ones. Everything else below rests on one or two small trials.
- Kidney stones (Moderate, for repeat stones). The landmark 5-year trial randomized 199 first-time calcium-stone formers to a high-water target (urine output above 2 L/day) or no treatment: recurrence fell to 12% versus 27%, roughly halving the risk, and stones that did recur took about a year longer to appear (39 versus 25 months).[16] A dose-response meta-analysis of 15 observational studies — 10 cohort and 5 case-control, no trials — covering 9,601 stone cases among 351,081 people puts a gradient on it — each additional 500 mL/day of water was associated with about 7% lower stone risk.[17] Borghi's trial arm aimed at more than 2 L/day of urine; current clinical guidance sets the bar a little higher, at 2.5 L/day for people who form stones. Borghi's result has not held at scale. The largest randomised test of hydration against stone recurrence coached 1,658 stone formers with low urine output through a fluid prescription, health coaching, text messaging and financial incentives, and found no benefit over guideline-concordant care: 154 of 826 in the intervention arm had a symptomatic recurrence over two years versus 165 of 832 controls, about one in five either way, with no difference in stone growth on end-of-study imaging — even though 24-hour urine volume genuinely did rise in the intervention arm.[18] That trial tested an adherence programme rather than a fixed water dose, and intake rose in both arms, so the contrast between them was diluted — which is why the older trial, run against no treatment at all, still counts. Read the stone evidence as secondary prevention only: the definitive systematic review found no randomised evidence at all that drinking more prevents a first stone, and graded the 1996 trial low-certainty.[19]
- Weight management (Weak). Three small trials of a roughly 500 mL premeal water preload reported greater weight loss than control, but the absolute effects are small, every one was run in middle-aged or older adults, and each has a catch. Forty-eight adults aged 55–75 on a reduced-calorie diet lost about 2 kg more over 12 weeks.[20] Eighty-four UK primary-care adults with obesity lost 1.3 kg more over 12 weeks — an advantage that shrank to 1.2 kg and lost statistical significance once demographics were accounted for, leaving a range from 2.4 kg lost to a slight gain — against a comparator asked to imagine a full stomach.[21] The third was not a weight-loss trial at all but a 12-month weight-loss maintenance trial in 40 already weight-reduced adults averaging 63 years, whose between-group comparison was null; its widely quoted "87% greater weight loss" comes from a separate model of self-reported daily weights and amounts to 0.67 kg lost against 1.00 kg regained.[22] A fourth trial, in adolescents and without the premeal timing, found nothing.[23] The "44–100% greater weight loss" figure in circulation is a relative framing from a review's narrative table, not a pooled estimate: when water-intake trials are actually pooled — seven randomised trials, only one of them a premeal preload and one of them the adolescent trial above — the effect on body weight disappears, at 0.33 kg lost with a range from 1.75 kg lost to 1.08 kg gained. The one hint of benefit that survives pooling is water displacing sugar-sweetened drinks, and that is displacement rather than hydration.[24] Two crossover studies do support a short-term satiety mechanism, specifically in older adults: a preload of 375 mL for women and 500 mL for men cut the next meal by about 58 kcal in healthy adults aged 60–80, an effect not seen in 21–35 year olds in the same study,[25] and a 500 mL preload cut breakfast intake by about 13% in 24 adults with obesity averaging 61 years.[26] A satiety effect at one meal is not the same as weight lost over months.
- Recurrent urinary tract infections (UTIs) in premenopausal women (Moderate, single trial). In premenopausal women with recurrent cystitis who habitually drank little, adding 1.5 L of water a day left them with an average of 1.7 episodes over a year against 3.2 in the control group — nearly halving both infections and antibiotic courses — through simple mechanical flushing.[27] That trial's own result was unambiguous, but it is one trial. Only one other trial in the review was null, and it measured urine cultures over four days rather than clinical episodes. The one clinical-episode comparison since put extra hydration behind both D-mannose and low-dose antibiotic prophylaxis in 75 premenopausal women — about one infection a year on hydration against 0.3 on D-mannose and 0.2 on antibiotics — though it had no untreated control, so it neither replicates nor refutes this result.[28] Read the original trial with a grain of salt for a second reason: it was funded and staffed by a bottled-water company, the kind of conflict of interest worth flagging.
- Migraine prophylaxis (Weak — and weaker than usually reported). Two trials tested it. The larger one improved migraine-specific quality of life but made no difference to the number of days with at least moderate headache; the smaller one, an 18-person pilot, was too small to detect anything either way. Dehydration is a plausible headache trigger, but "drink more water for migraines" is not something these trials established.
- Glycemic control and blood pressure (Weak, single trials). One 40-person trial in type 2 diabetes found a large drop in fasting glucose from premeal water; a second trial in adults with mildly raised glucose found none. In 20 healthy volunteers, an extra 30 mL/kg/day for two weeks — roughly 2 L more — raised daytime ambulatory mean arterial pressure from 89.0 to 91.4 mmHg, a rise rather than a fall, while night-time pressure did not move and a single self-report item about vertigo improved.[29] Read that as a reason not to assume more is always better, not as a benefit.
One important counterweight keeps the "more is always better" story honest. The most rigorous kidney trial was null: 631 adults with moderate chronic kidney disease (CKD) coached to drink about 1 L/day more showed no slower decline in kidney function after a year — the difference in filtration rate was −0.3 mL/min/1.73 m², with a range of −1.8 to 1.2 that comfortably straddles zero — even though their copeptin (the vasopressin surrogate) did drop.[30] That bracketed range is a confidence interval: it's where the true effect most plausibly lies, and when it spans zero (or, for a ratio, crosses 1.0), the result is compatible with no effect at all. A second, longer trial says the same thing from another angle: three years of prescribed high water intake in 184 adults with polycystic kidney disease did not slow kidney growth — about 6.8% per year against 7.8% on ad libitum drinking, a difference compatible with none — and, unlike the chronic kidney disease trial, copeptin did not fall at all despite a genuine separation in urine concentration.[31] That trial was part-funded by a bottled-water company's research arm, and four authors of the chronic kidney disease trial declare ties to the same company — the same conflict flagged above for the urinary-tract-infection trial. So the serum-sodium and cohort signals above are best read as associations that hydration may improve, not proof that drinking more rescues an already-diseased organ.
How much should you actually drink?
Evidence: Strong that there is no single correct volume and that turnover is not a drinking target; Weak for the older-adult guidance.
There is no universal volume target. The "8 × 8 oz" rule is folklore, not evidence. The best objective evidence comes from a stable-isotope study of 5,604 people aged 8 days to 96 years across 23 countries, which measured how fast water actually cycles through the body: turnover peaked around age 30 at about 4.3 L/day in men and 3.4 L/day in women, fell to 3.1 and 2.8 L/day past 70, and ranged from roughly 1 to 6 L/day between individuals, driven by climate, lean mass, athletic status, and pregnancy.[32]
Crucially, turnover is not a drinking target (Strong). About 10% of it is water your own metabolism makes, and another few per cent is absorbed through the lungs and skin, so food and drink together account for roughly 85% of the total. How that 85% splits between eating and drinking is genuinely unresolved — the study's authors call it an open question and note that dietary surveys, which put the food share anywhere from 20% to 50%, probably underestimate it (Weak). Official reference values assume the low end: European authorities set adequate total water intake at 2.5 L/day for men and 2.0 L/day for women, and only for moderate temperatures and moderate activity;[33] the US figures are higher at 3.7 and 2.7 L/day.[34] Both include water from food and neither is an individual prescription, but they were built differently: the US value is simply the median total water intake in a national survey, while the European one is back-calculated from a target urine concentration on the assumption that drinks supply about 80% of the total.
Plain water is only one part of that total. In an analysis of three cycles of the US national nutrition survey (2011–2016), plain tap and bottled water averaged about 1.1 L/day against roughly 2.7 L/day of total dietary water, and only about 40% of the sample met the US adequate-intake reference — an analysis co-authored by beverage-industry staff, and one that says nothing about where anybody's serum sodium sat.[35]
Practical biomarker-based check:
- Thirst rarely activated during normal daytime activity.
- Urine consistently pale yellow / nearly colorless through the day (first morning urine excepted). This check is for healthy midlife adults only: the European geriatric nutrition guideline carries a Grade A instruction not to judge hydration from urine colour or specific gravity past 65, because a diagnostic-accuracy review found none of the simple bedside signs reliable in older people — measured or calculated blood osmolarity is the reference standard there.[36]
- No frequent unexplained mild headache, lethargy, or afternoon brain fog — commonly attributed to, though not established as, mild dehydration.
If those are off, raise intake. Most healthy midlife adults will land somewhere between 2.0 and 3.0 L/day total fluids (food included); athletes, hot climates, and people on a glucagon-like peptide-1 (GLP-1) agonist need more. The usual justification for that on a GLP-1 agonist does not survive the only randomised test of it: in healthy volunteers, three weeks of dulaglutide left thirst perception and blood sodium unchanged, and although people drank a little less, urine output fell further still, so fluid balance did not shift in a dehydrating direction.[37] The real reasons to watch intake are the gastrointestinal side effects, which can cause genuine fluid loss through a bout of nausea or diarrhoea. No guideline, drug label or trial sets a fluid volume for people on these drugs, so treat the commonly quoted 2.5–3.5 L/day as a rule of thumb rather than a prescription.
One popular claim to temper: there is very little evidence either way that drinking extra fluid, on its own, relieves chronic constipation in people who are already well-hydrated. The UK dietetic guideline could find only a single trial of water supplementation, and concluded that whether drinking beyond habitual intake helps is still unresolved rather than that it does not — in that trial both arms ate the same fibre, so the fluid was the only variable, but one small trial cannot settle it either way (Weak).[38]
Older adults are the group where thirst becomes an unreliable guide: aging blunts both the thirst signal and the kidney's ability to concentrate urine, so low-intake dehydration is common. A meta-analysis of 61 studies in non-hospitalised people over 65, 44 of them pooled and covering 22,398 participants, put the prevalence of low-intake dehydration at about a quarter — roughly one in five community-dwelling older adults and one in three long-term-care residents, though the study-to-study spread was so wide that these are ballpark figures rather than precise ones.[39] It carries weight in hospital. A 2026 systematic review of 19 studies covering 14,398 hospitalised adults aged 65 and over puts dehydration on admission at roughly a quarter to a third — 23% pooled, 35% counting patients directly — and found that it barely improves during the stay, though the studies varied so widely that the range matters more than the point estimate.[40] In one prospective study of 200 emergency admissions aged 65 and over, 37% arrived dehydrated and 62% of those were still dehydrated two days later; adjusted for age, sex, comorbidity, illness severity, frailty and nutritional risk, the dehydrated patients were around six times more likely to die in hospital — an alarming figure, but one resting on just 14 in-hospital deaths, with a plausible range running from a 64% increase to more than twentyfold (hazard ratio 6.04, 95% CI 1.64–22.25).[41] For anyone past midlife, that argues for drinking on a rhythm rather than waiting to feel thirsty — and, as noted above, for not leaning on the urine-colour check past 65; the same guideline also rules out skin turgor, dry mouth and weight change as hydration checks. It gives the one concrete volume target in this area: at least 1.6 L/day of drinks for older women and 2.0 L/day for older men. It also notes that no drinking intervention has yet been shown to prevent or treat low-intake dehydration (Weak).[42]
Thirst is also the guardrail against the opposite error. Overdrinking during endurance exercise causes exercise-associated hyponatremia — dangerously dilute blood sodium — and it is driven by drinking too much plain fluid, not by losing salt, so sports drinks do not prevent it. The international consensus on the condition, which counts at least 14 athlete deaths since 1981, is blunt: drinking to thirst prevents both dehydration and hyponatremia, and the "drink before you're thirsty" advice is wrong for most people. The panel makes that a strong recommendation but grades the evidence behind it low, and flags exceptions — a dry mouth, a very high salt intake and other non-osmotic drives can all make thirst an unreliable guide (Moderate).[43]
Drink it slowly: bolus vs. metered
Evidence: Weak — one small trial with a clean mechanism behind it. Harmless advice, but don't mistake it for a settled finding.
A frequent failure mode is chugging — slamming a liter of water in five minutes to "catch up." It probably doesn't work the way people think.
In an environmental-chamber study, eight healthy adults were dehydrated by 2% of body weight through exercise at 35 °C, then rehydrated to identical volumes by either (1) drinking the full amount in one hour or (2) sipping 12.5% every 30 minutes over four hours.[44]
| Pattern | Urine output | Hydration efficiency |
|---|---|---|
| Bolus (1 hour) | 700 mL | 55% |
| Metered (4 hours) | 420 mL | 75% |
Why: a sudden flood drops plasma osmolality fast, the hypothalamus suppresses AVP, the kidneys clear the surplus, and less of it is retained. That chain is the standard physiological explanation rather than something this trial demonstrated: it measured urine volume and net fluid balance only, with no measurement of vasopressin, plasma concentration, or where in the body the water ended up. Metered intake keeps AVP steady and lets the gut and vasculature distribute the water properly.
Two caveats that the usual retelling of this study drops. It has eight participants, and it tested post-exercise rehydration in a heat chamber rather than ordinary daytime drinking. A third caveat matters more: an earlier randomised crossover of the same size pointed the other way. Eight trained cyclists dehydrated by 3% of body weight and given a carbohydrate-electrolyte drink passed more urine early on the fast schedule, but that reversed by hours five and six, plasma volume and fluid balance were restored faster on the fast schedule, and the difference had vanished by six hours.[45] The drinks differed — plain water versus a carbohydrate-electrolyte solution, itself a major determinant of retention — so the two do not cancel, but the metered-versus-bolus literature is genuinely split. The physiology generalizes well enough to justify the habit; the specific 55%-versus-75% numbers should not be treated as a law.
Practical: keep a glass at arm's reach and refill on a 30–60 minute rhythm. Avoid a large glass right before bed if you already wake to urinate. Limiting evening fluid is a standard first-line measure for nocturia, which is itself linked to falls and fractures, low mood and cognitive impairment.[46] No trial has compared a bedtime glass against the same water taken earlier for sleep quality, so treat this as sensible rather than demonstrated (see Sleep for context on sleep continuity).
What's actually in your water
The most consequential water-quality decisions for a longevity-conscious adult are no longer about E. coli — they're about chronic exposure to two classes of pervasive synthetic contaminants.
Microplastics and nanoplastics
Evidence: Strong that they are present in drinking water; Moderate that they reach human tissue; Weak on what any given dose does to health.
Microplastics (<5 mm) and nanoplastics (<1 µm) — polyester, polyethylene, polypropylene, polystyrene — are now ubiquitous in tap and bottled water. An earlier survey of 259 bottles across 11 major brands found signs of microplastic contamination in 93% of them, averaging ~325 particles/L,[47] but a 2024 study using a laser-imaging technique able to see far smaller particles pushed the count two to three orders of magnitude higher: bottled water averaged roughly 240,000 particles per liter, about 90% of them nanoplastics — small enough that they could cross biological barriers.[48] Treat that headline number as directional, not exact: a published critique argues the study's procedural blanks were as contaminated as the bottled water and that the dominant polymer detected was nylon-like polyamide rather than the bottle's polyethylene terephthalate (PET), pointing to contamination or identification problems — making the absolute count unreliable even if the qualitative conclusion holds; the authors dispute the critique.[49] Either way, bottled tends to exceed tap, and the packaging is part of the reason: in the 259-bottle survey the most common confirmed polymer was polypropylene, the cap material, and the authors concluded the contamination comes at least partly from packaging and bottling.
Microplastic and nanoplastic fragments have been recovered at autopsy from human brain, liver and kidney tissue, and the brain particles were largely nanoscale shards.[50] Separate groups have reported them in placenta[51] and in lung tissue, where the authors attribute the route to inhalation rather than swallowing.[52] A further group reported plastic particles in the blood of 22 donors, a result other researchers have contested.[53] What follows from those detections is far less certain. The evidence that particles cross the gut lining, that tissue-bound particles drive local chronic inflammation, or that they ferry endocrine-disrupting plasticizers such as bisphenols and phthalates into the body comes from animal and cell experiments, not from people; a systematic review of the field concludes that direct human health evidence is still scarce.[54]
PFAS — "forever chemicals"
Evidence: Strong on hazard and regulatory direction; Moderate on the risk at everyday exposure levels.
PFAS (per- and polyfluoroalkyl substances) are virtually indestructible carbon-fluorine compounds used since the 1940s in non-stick cookware, water-repellent textiles, and firefighting foams. They accumulate in groundwater, food chains, and human serum.[55]
The evidence has hardened. In late 2023 the World Health Organization's cancer agency reclassified perfluorooctanoic acid (PFOA) as a Group 1 human carcinogen — its highest certainty tier, up from "possibly carcinogenic" a decade earlier — and rated the related compound perfluorooctanesulfonic acid (PFOS) as possibly carcinogenic; that grades hazard, not the risk at everyday exposures.[56] The heavily studied "C8" community cohort near a contaminated US chemical plant found testicular-cancer risk climbing across exposure quartiles, with a weaker and statistically non-significant gradient for kidney cancer — which is why the cancer agency graded the human cancer evidence for PFOA only "limited", resting its Group 1 call on animal and mechanistic data instead.[57] And in a Faroe Islands birth cohort, each doubling of children's blood PFAS level went with roughly half the antibody concentration against tetanus and diphtheria vaccines — the immune endpoint European regulators used to set safe-intake limits (Moderate).[58] Beyond cancer and immunity, chronic exposure links to thyroid disruption, abnormal blood lipids, and reproductive harm; a quasi-experimental study of every New Hampshire birth from 2010 to 2019, comparing mothers on wells downgradient of PFAS-contaminated sites with those upgradient, found sharply higher infant mortality, very preterm birth and very low birth weight, and priced the lifetime social costs of those three outcomes — medical care, long-term health effects and reduced lifetime earnings — at roughly $8 billion for each year's birth cohort across the contiguous United States. That is an infant-health figure extrapolated from 11 states, not a total PFAS cost of illness.[59]
Regulators moved accordingly: the US Environmental Protection Agency's 2024 drinking-water rule set enforceable limits of 4 nanograms per liter (parts per trillion) for both PFOA and PFOS, with a health-based goal of zero. As of mid-2026 that rule is being rewritten rather than abandoned — the agency has proposed keeping the 4 ng/L limits for PFOA and PFOS while offering water systems that apply for it a two-year exemption from the April 2029 compliance date, out to April 2031 — systems that do not apply keep 2029, and those measuring at or above 12 ng/L would have to install interim controls in the meantime. It has separately proposed rescinding the limits for four other PFAS compounds on legal and procedural grounds rather than on any reassessment of their health risk.[60] Those four limits remain enforceable law: a federal appeals court refused in January 2026 to vacate them at the agency's request, and the case has not been decided. The enforcement timeline is slipping; the toxicology has not changed. Outside the US the benchmark is different in kind: since January 2026 the EU's recast Drinking Water Directive has applied a limit of 100 nanograms per litre for the sum of 20 named PFAS, or alternatively 500 ng/L for PFAS as a whole, with no single-compound ceiling and no zero goal — roughly 25 times more permissive as a sum than the US limit on PFOA and PFOS individually.[61]
This places water filtration in the same risk category as the modifiable environmental exposures already discussed in Ultra-processed food and Environmental toxins: the dose-response is hard to detect at the individual level, but the population signal is consistent for immune endpoints, while the human cancer evidence is rated only limited. It is weaker elsewhere: the C8 effect sizes come from a partially overlapping, high-exposure population, and the national-survey mortality evidence is weak and partly withdrawn — the most-cited positive result, from a US health survey, was retracted in 2025.[62] The clearer mortality signals now come from heavily contaminated communities rather than the general population: a Swedish register study of 46,553 adult residents of one town found cardiovascular mortality about 15% higher among those ever supplied by its contaminated waterworks.[63]
The older waterborne carcinogens: arsenic, nitrate, and chlorination byproducts
PFAS and microplastics grab the headlines, but three long-recognized contaminants matter for a longevity-conscious adult and strengthen the case for filtration:
- Arsenic (Moderate-to-strong at high exposure; contested at low). Reviewing 30 years of evidence, exposure at 10 µg/L — the WHO and US regulatory limit — was modelled as roughly doubling bladder-cancer risk — but that is an extrapolation from a literature dominated by far higher exposures, and the same authors' robustness estimate at 10 µg/L has a confidence interval that includes no effect at all.[64] A 2026 pooled analysis of 62 studies found the dose-response gradient only above 10 µg/L, not at or below it, and put bladder-cancer risk about 60% higher in the top exposure tertile versus the bottom.[65] The excess is clear at high concentrations; at typical low exposures the risk is genuinely uncertain, because exposure is harder to classify accurately at the bottom of the curve. Mostly a concern for private wells in geologically arsenic-rich areas.
- Nitrate (Moderate). In a Danish register study that estimated exposure for 2.7 million adults and analysed 1.7 million of them in detail, colorectal-cancer risk was about 16% higher in the most-exposed group, with a signal emerging around 4 mg/L of nitrate ion — well below the 50 mg/L regulatory standard.[66] That threshold has not held up. Two 2022 meta-analyses of the drinking-water literature found no clear association with colorectal cancer,[67][68] and a 2025 Danish cohort of 54,610 people with full individual dietary data — which largely answers the confounding objection — found nothing for colorectal cancer overall, but about 52% higher colon-cancer risk above roughly 9 mg/L of tap-water nitrate.[69] Nitrate is largely an agricultural-runoff problem.
- Trihalomethanes (Moderate). These chlorination byproducts, present in virtually all disinfected tap water, were associated with about 44% higher odds of bladder cancer among men at the highest exposures in a pooled analysis of six case-control studies — and with no detectable increase among women, though the data cannot rule out a modest one.[70] A 2025 dose-response meta-analysis covering 5,860 bladder-cancer cases has since gone further: it put risk about 33% higher at the highest exposures, statistically significant above 41 µg/L — below both the US and EU regulatory limits — and found a second site the older analysis did not, with colorectal-cancer risk about 15% higher.[71] This is a real trade-off of chlorination, not an argument against it — disinfection prevents far more disease than it plausibly causes — but it is one more reason a good filter earns its place.
Reverse osmosis handles all three; a simple activated-carbon filter substantially reduces trihalomethanes. Private wells are the exception worth checking: among 156 Maine households with arsenic treatment, mostly point-of-use reverse osmosis, about one in five still delivered water above the 10 µg/L limit, and failure was more likely the higher the share of arsenic present in its uncharged form and where the owner rather than the vendor maintained the unit.[72]
Filtration: RO vs. UF, and the demineralization trap
Evidence: Strong on what each technology removes; Moderate on the health consequences of removing minerals.
Two filtration technologies are realistic for the home — reverse osmosis (RO) and ultrafiltration (UF) — and they make different trade-offs.
| Tech | Pore size | Microplastics & pathogens | PFAS / dissolved chemicals | Heavy metals | Mineral retention |
|---|---|---|---|---|---|
| Reverse osmosis (RO) | ~0.0001 µm | Yes | Yes (highly effective) | Yes | No (strips ~94% of calcium, ~97% of magnesium) |
| Ultrafiltration (UF) | ~0.02 µm | Yes | No (passes through) | Minimal | Yes |
RO removes essentially everything — microplastics, pathogens, heavy metals, and the dissolved-chemical contaminants UF cannot touch. The price is that it strips out the beneficial minerals too. A field survey of community RO stations measured average rejection of 93.5% for calcium and 97.4% for magnesium, with total water hardness down 95.8%.[73]
UF keeps minerals intact but lets PFAS straight through, because it removes particles rather than dissolved solids. For most modern municipal water sources, that's a hard limitation.
Why demineralized RO water is its own problem
Evidence: Moderate for the mineral-balance and food-leaching effects; Weak for long-term skeletal consequences.
The case against habitually drinking demineralized water — reverse osmosis without remineralization, or distillation — is set out in a background chapter of a World Health Organization volume on nutrients in drinking water. It is a signed author contribution rather than WHO policy: the volume's front matter states that the named authors alone are responsible for the views expressed. The volume's expert-consensus chapter stops short of any consumer warning and addresses itself to suppliers, recommending that demineralized and corrosive water be stabilized with additives that restore calcium and magnesium.[74] Three problems, in descending order of how well established they are:
- Disturbed water and mineral balance. This is the part the chapter calls adequately demonstrated. In human volunteer experiments, water with total dissolved solids below about 100 mg/L raised urine output by almost 20% on average, raised serum sodium, lowered serum potassium, and increased the loss of sodium, potassium, chloride, calcium and magnesium from the body — though those experiments trace to an unpublished 1980 working document that reports no sample sizes or designs. (An older rat experiment suggesting that very low-mineral water damages the intestinal lining was not reproduced: a later 14-day rat study found no erosion, ulceration or inflammation in the oesophagus, stomach or jejunum, and the chapter concludes the data have not unambiguously demonstrated a direct effect on the gastrointestinal lining — while still reporting increased gastric acid secretion and altered stomach muscle tone.)
- Nutrient leaching from food. Demineralized water acts as a solvent. Cooking vegetables, meat, cereals or pasta in soft water was found to pull up to 60% of their calcium and magnesium into the cooking water — and more for some trace elements, around 66% of copper, 70% of manganese and 86% of cobalt — all lost if the water is discarded. Demineralized water sits at the extreme end of that range, though the underlying studies are old and report no sample sizes. Long-term consumption may also draw on the body's own mineral reserves; animal work found reduced skeletal ossification in rat foetuses whose mothers were given distilled water, and two cohorts of women in South Siberian cities supplied with very low-calcium, low-magnesium water showed more osteoporosis on X-ray absorptiometry than those in higher-mineral cities — but no effect sizes are reported for any of it, so the human evidence is suggestive rather than settled.[75]
- Higher heavy-metal uptake. Calcium and magnesium compete with lead and cadmium for absorption in the gut and can form poorly absorbed compounds with them. Strip them out and that modest protective effect goes with them.
The cardiovascular case for "hard" water
Evidence: Moderate — one large cohort, pointing the same way as fifty years of ecological hard-water studies, but observational, confined to stroke, and set against a WHO expert review that discounts those ecological studies on design grounds and reports that the cohort studies of water hardness found no association at all.
Drinking water has historically been a significant source of bioavailable calcium and magnesium. The Swedish Mammography Cohort — 26,733 postmenopausal women followed for 16 years — compared municipal water high in these minerals (magnesium averaging 10 mg/L) against low (averaging 5 mg/L) and found about 13% lower ischemic stroke risk and 22% lower hemorrhagic stroke risk. When the two minerals were separated, only magnesium carried the association, and it carried it strongly: women in the top third of water magnesium had about 31% lower ischemic stroke risk than the bottom third (HR 0.69) — a separate tertile analysis, not the 10-versus-5 mg/L contrast above.[76] Note what the same study did not find — no significant effect on heart attacks (HR 0.93, 0.85–1.02, a range that includes no effect). This is a stroke finding, and specifically a magnesium finding, not a general "hard water protects the heart" result.
This is why RO + remineralization is the protocol of choice if your water source is suspect for PFAS, and why straight RO without remineralization is a poor default.
Remineralization, in order of evidence
Evidence: Moderate — the targets are researchers' proposed floors rather than guideline values; which delivery method best hits them has not been trialed head to head.
There is no World Health Organization guideline for minimum calcium or magnesium in drinking water — its 2009 expert review says so outright — but that same review notes that where hard-water studies found a significant cardiovascular benefit, it generally appeared at magnesium concentrations of about 10 mg/L and above.[77] A background chapter in WHO's earlier volume on nutrients in drinking water collects what individual researchers have proposed as floors for demineralized supplies: about 10 mg/L magnesium, 20 mg/L calcium (the 1980 WHO-commissioned report had proposed 30), and total dissolved solids no lower than 100 mg/L; the proposed optima are higher, about 20–30 mg/L magnesium and 40–80 mg/L calcium.[78] Those are reasonable targets to aim at, not international standards — and national rules across the European Union remain inconsistent and largely not based on current research.[79] That magnesium floor is the same concentration as the protective tier in the Swedish cohort.
- Inline remineralization cartridge as a final RO stage. Calcium carbonate and magnesium-oxide media slowly dissolve into the aggressive, slightly acidic RO water, restoring calcium and magnesium and bringing pH from ~5.5 to neutral. Most consistent and convenient.
- Ionic trace mineral drops. Concentrated ionic trace mineral solutions (often sourced from inland seas like the Great Salt Lake) added per pitcher/bottle. High bioavailability, but requires daily compliance.
- Pinch of unrefined salt (Himalayan, sea). Popular DIY, but flawed — to deliver clinically meaningful magnesium and calcium via salt alone, you would need roughly 30 g of salt a day, six times the WHO's 5 g upper limit. Don't rely on salt as your remineralization strategy.
Electrolyte powders: when you actually need them
Evidence: Moderate — the case against routine daily use is solid; the exact threshold above which they help is a rule of thumb, not a finding.
Healthy office workers are heavily marketed a daily electrolyte packet to ward off "brain fog" and "cramps." For most adults, there is nothing behind that.
The most informative evidence comes from the opposite extreme. Researchers followed 266 competitors through week-long desert ultramarathons and measured blood sodium after the race in 174 of them. Forty-one had a sodium disturbance — but 30 of those were concentrated blood from dehydration, not the dilute blood that electrolyte marketing targets — and neither the rate nor the total amount of supplemental sodium taken predicted who was affected. Hot races carried roughly nine times the odds of the high-sodium form. Supplement use was self-selected rather than assigned, so this is an observational signal rather than a settled causal answer.[80] What did protect people was training volume, avoiding overhydration, and cooler conditions. A packet does not rescue an ultrarunner in the heat; there is little reason to expect more of it at a desk.
The conventional sports-nutrition rule of thumb is that added electrolytes start to matter somewhere past 60–90 minutes of sustained exertion, or sooner in heavy heat with visible salt loss in sweat. Treat that as a rough guide rather than a validated cutoff — no trial has established a specific threshold. Below it, plain mineralized water is sufficient.
The downsides of routine daily use:
- Many commercial powders carry substantial added sugar (fructose/glucose, used to drive sodium co-transport across the gut). Daily use can push past WHO/American Heart Association added-sugar limits.
- Excess potassium loading is risky in anyone with kidney disease or on drugs that cut renal potassium excretion, such as angiotensin-converting-enzyme (ACE) inhibitors, angiotensin-receptor blockers (ARBs) and spironolactone (the list also includes NSAIDs and trimethoprim).[81] The 2025 US hypertension guideline recommends potassium salt substitutes and potassium supplements for blood pressure but carves out chronic kidney disease and drugs that reduce potassium excretion as cases where serum potassium should be monitored, so the practical rule is to check with a doctor before adding a potassium-loaded product.[82]
- Routine high-dose sodium is exactly the wrong direction for most adults given the population sodium burden.
When electrolytes do help: long Zone 2 sessions, hot-weather endurance work, the sauna recovery context, gastrointestinal illness with significant losses, and on a GLP-1 receptor agonist where appetite suppression and gastrointestinal side effects can leave the user mildly volume-depleted.
Trace minerals: a possible, unquantified gap
Evidence: Weak-to-Moderate — the biochemistry is settled; the claim that soil depletion has made subclinical deficiency widespread in well-fed populations is not well quantified.
While the average adult does not need supplemental sodium or potassium, subclinical deficiencies of trace minerals — zinc, copper, selenium, manganese, iodine, iron — are often said to be common, blamed on industrial-farming soil depletion and ultra-processed food intake, but how common they are in well-fed adults has not been well measured. Several of them sit at the catalytic core of the body's own antioxidant enzymes: the enzyme that neutralizes superoxide radicals requires copper, zinc and manganese, and the one that clears peroxides requires selenium. Run short and oxidative stress rises directly.
Practical: a Mediterranean-pattern diet (Dietary patterns) plus moderately mineralized water covers most of this. Ionic trace mineral drops are a reasonable inexpensive insurance policy for adults relying on RO water.
For the dedicated magnesium story, including how supplemental magnesium relates to blood pressure and sleep on top of waterborne intake, see Magnesium.
Functional waters: what the evidence actually says
A scientific look at the four most-marketed water modifications.
Molecular hydrogen water — preliminary evidence, plausible mechanism
Evidence: Weak — real biology, real small trials, no hard outcome data.
Water supersaturated with dissolved hydrogen gas (H₂). Unique among the "functional water" category in that it has accumulating peer-reviewed clinical evidence rather than purely commercial backing. H₂ is a small, neutral molecule that crosses cell membranes and the blood-brain barrier and acts as a selective antioxidant — mopping up the most destructive reactive oxygen species while leaving intact the milder ones the body uses for signaling. A 2023 review counted 81 registered trials and 64 published human studies across cardiovascular, respiratory, neurological and other indications; note that its authors declare a patent application on hydrogen-release technology.[83]
Small trials, mostly in people with metabolic syndrome, have shown reductions in total and oxidized low-density lipoprotein (LDL) cholesterol and in the standard inflammatory markers over 10 weeks to 6 months.[84] A separate systematic review of 25 studies reached a deliberately deflationary verdict — its title asks whether hydrogen water is "extra healthy or a hoax" — concluding that the preliminary results are encouraging but that larger, more rigorous trials are needed before anything can be claimed.[85] Honest verdict: the most plausible "functional water" with real biology behind it; not yet supported by hard outcome data.
Alkaline water — claims fail under physiology
Evidence: Strong (against any benefit); Caution in chronic kidney disease.
The marketing premise — that high-pH water "neutralizes systemic acid" — is incompatible with how the body works. Arterial pH is held at 7.35–7.45 by powerful renal and respiratory buffers; ingested fluid is neutralized by gastric acid at pH 1.5–3.5 long before any "alkalinity" reaches the blood. A systematic review found zero significant differences in gut microbiota, blood parameters, urine pH, or fitness performance versus plain mineral water.[86] Strongly alkaline water (pH above about 9.8) is worth avoiding in chronic kidney disease, where a reduced ability to excrete potassium makes any added potassium load a hazard — a theoretical concern rather than a documented series of cases, but an easy one to sidestep. Skip.
"Structured water" / hexagonal water / H₃O₂ — pseudoscience
Evidence: Strong (against).
Sold as water molecules permanently rearranged into a stable hexagonal lattice that hydrates better. Physicochemically impossible — liquid water hydrogen bonds break and reform on picosecond timescales. The legitimate Pollack "exclusion zone" research describes a transient interfacial gel layer at hydrophilic surfaces, not bulk bottled water. The structure cannot be packaged or sold; "structured water" funnels and wands are snake oil.[87]
Marine plasma / Quinton water — unsubstantiated
Evidence: Weak (no support), with a documented safety action.
Microfiltered deep-sea water sold for "stress" and "nervous system" benefits. The human evidence amounts to one small, manufacturer-linked randomised trial of a deep-seawater mineral extract in 75 people with impaired glucose tolerance, which was null on every hard glycemic endpoint and tested neither a branded marine-plasma product nor the "stress" and "nervous system" claims;[88] the positive cited data are otherwise extrapolated from cell and animal studies using laboratory-grade deep-ocean minerals, not commercial product. In 2006 Health Canada told consumers not to use two Ocean Plasma products from a single Montreal manufacturer: they were unapproved and contained unacceptable amounts of aerobic bacteria, which can cause fever, vomiting, diarrhoea and abdominal pain. The advisory also recorded that no adverse reactions had ever been reported to the agency, and no comparable action against marine-plasma products has followed in the twenty years since — so read it as a licensing and manufacturing failure at one company rather than as evidence that the category makes people ill.[89] Use mineral drops or mineralized RO water instead.
A practical water protocol
For a healthy midlife adult who wants to extract the available longevity benefit:
- Drink steadily, not in bolus — a glass on a 30–60 minute rhythm during waking hours; final glass well before bed to avoid sleep disruption.
- Aim for pale-yellow urine and rare thirst as the daily check; volume usually lands at ~2.0–3.0 L total fluids.
- Filter for what's actually in the supply. Match the filter to what your utility's report shows. Carbon pitchers remove a large but variable fraction of PFAS; reverse osmosis removes more, and also removes fluoride and nitrate, though about one in five treated private wells in a Maine survey still tested above the arsenic limit — most often where the householder rather than the vendor maintained the unit — and it strips calcium and magnesium unless you add a remineralization stage. If you do, roughly 10 mg/L magnesium and 20 mg/L calcium in the finished water are sensible floors — researchers' proposals, not WHO standards.
- Avoid daily bottled water as a primary source. Microplastic concentration is typically higher than tap, partly from the packaging itself. A reusable stainless / glass bottle filled from a properly filtered tap is materially better.
- Skip electrolyte powders unless training beyond about an hour, in heat, ill, or on a GLP-1. When you do need them, prefer low/no-sugar formulas.
- Trace minerals from food, plus mineralized water, is the right baseline. Drops are reasonable on RO setups.
- Functional waters: hydrogen water is the only one with real evidence and even that's preliminary. Alkaline, structured, marine-plasma — pass.
- If you form kidney stones, higher fluid intake is still standard care — but the largest trial of it was null, so read the section above before treating it as a guarantee.
Further reading
The thirteen that carry the most weight; everything else is linked inline.
- Dmitrieva NI et al. Middle-age high normal serum sodium as a risk factor for accelerated biological aging, chronic diseases, and premature mortality. EBioMedicine 2023.[90]
- Rabinowitz J et al. Risk of hypertension and heart failure linked to high-normal serum sodium and tonicity (Leumit cohort). Eur J Prev Cardiol 2025.[91]
- Hakam N et al. Outcomes in randomized clinical trials testing changes in daily water intake: a systematic review (18 trials; only weight loss and stones backed by more than one trial). JAMA Netw Open 2024.[92]
- Borghi L et al. Urinary volume, water and recurrences in idiopathic calcium nephrolithiasis: a 5-year randomized prospective study. J Urol 1996.[93]
- Xu C et al. Self-fluid management in prevention of kidney stones: dose-response meta-analysis. Medicine (Baltimore) 2015.[94]
- Clark WF et al. CKD WIT — coaching to increase water intake in chronic kidney disease (null on kidney function). JAMA 2018.[95]
- Yamada Y et al. Variation in human water turnover (doubly labeled water, 23 countries). Science 2022.[96]
- Hew-Butler T et al. 3rd International Exercise-Associated Hyponatremia Consensus. Clin J Sport Med 2015.[97]
- Parkinson E et al. Low-intake dehydration prevalence in non-hospitalised older adults. Clin Nutr 2023.[98]
- Qian N et al. Rapid single-particle chemical imaging of nanoplastics by SRS microscopy. PNAS 2024.[99] — read alongside its methodological critique.[100]
- Zahm S et al. Carcinogenicity of perfluorooctanoic acid and perfluorooctanesulfonic acid (IARC). Lancet Oncol 2024.[101]
- Kozisek F. Health risks from drinking demineralised water, in WHO, Nutrients in Drinking Water, 2005 — a signed background chapter, and the source of the cooking-loss figures.[102]
- Helte E et al. Calcium and magnesium in drinking water and risk of myocardial infarction and stroke. Am J Clin Nutr 2022.[103]