The unglamorous evidence — what actually has mortality data

What actually decides this outcome, in order of size

Four variables carry almost all of the causal human evidence for how long somebody lives, and the reason this page exists is that they are boring. Ranked by the quality of the evidence behind them rather than by how interesting they are:

1. Atherogenic particle number, where the relationship is causal and dose-dependent. Meta-analysis of 170,000 participants across 26 randomized trials found that more intensive lipid lowering produced a further 15% reduction in major vascular events (95% CI 11 to 18, P<0.0001) for an additional 0.51 mmol/L of LDL cholesterol at one year Baigent 2010. Not an association across a cohort: a further reduction for a further lowering, inside randomized trials, which is what causal means.
2. Blood pressure, where the target itself was tested. Randomly assigning a systolic target below 120 rather than below 140 produced a primary outcome rate of 1.65% per year against 2.19%, a hazard ratio of 0.75 (95% CI 0.64 to 0.89, P<0.001) Wright 2015. The trial was stopped early for benefit, which is itself a data point about effect size.
3. Muscle and strength, measured as grip. In 502,293 UK Biobank participants followed a mean 7.1 years, during which 13,322 died, grip strength was associated with cardiovascular, respiratory and cancer outcomes and with all-cause mortality Celis-Morales 2018. This is observational and grip is a proxy, and it is still the strongest signal any modifiable body-composition variable has.
4. Glycemia, which is the slowest to move and the earliest to start. Non-enzymatic glycation of long-lived proteins is cumulative and effectively irreversible, so the exposure that matters is the integral over decades rather than this quarter's number.
5. Activity, where the first hour is worth the most. The pooled dose-response analysis of leisure-time activity found the steepest part of the curve at the bottom Arem 2015, which means the person with the most to gain from this page is the one currently doing nothing.

The order to run these in, and what has to be true first

Measure the four, treat the two that have drugs, train the one that does not, and buy the supplements last with their ceilings attached.

1. The four numbers, one draw and one cuff. ApoB (Apolipoprotein B) rather than the Lipid Panel (Cholesterol, HDL, LDL, Triglycerides) alone, because ApoB counts one particle per molecule and does not care about particle size; Lipoprotein(a) — Lp(a) once in a lifetime, because it is largely genetically determined and no lifestyle lever moves it; HbA1c (Hemoglobin A1c) with Fasting Insulin; and a validated home blood pressure average rather than a single clinic reading.
2. Rosuvastatin and Ezetimibe are on the same shelf because they attack the same number from two ends. A statin inhibits HMG-CoA reductase, which depletes hepatic cholesterol and upregulates LDL receptor expression; ezetimibe blocks NPC1L1-mediated intestinal absorption, which raises the hepatic demand the statin is already amplifying. That is why the combination is more than additive on arithmetic and why the trial evidence supports the pairing.
3. Icosapent Ethyl is the one omega-3 product with a cardiovascular outcome behind it. High-dose purified EPA reduced events in a statin-treated population with raised triglycerides Bhatt 2019, and it is not interchangeable with a gram of mixed fish oil, which is the distinction the whole category blurs.
4. Telmisartan, Losartan, Lisinopril and Nebivolol are four different mechanisms for one number. The ARBs block angiotensin II at the AT1 receptor, the ACE inhibitor removes its production and also raises bradykinin, which is where the cough comes from, and nebivolol is a beta-1 selective agent with nitric oxide mediated vasodilation. The class matters less than the achieved number does Wright 2015.
5. Creatine and Whey Protein (RecoveryPro) are the muscle half, and they only work attached to training. Protein supplementation adds a bounded increment to resistance-training gains and the increment shrinks as habitual intake rises Morton 2018. Neither does anything without the stimulus, which is the part of this page that cannot be purchased.
6. Vitamin D, Omega-3 (Fish Oil), Vitamin K2 Complex and Magnesium are last, and honestly last. They are adequacy insurance with real mechanisms and no mortality data at the doses sold, which is a different sentence from saying they do nothing.

What gets bought for this that cannot move it

A nutritional dose of omega-3 does not buy a pharmacological outcome, and the trial that shows it is large. In 15,480 adults with diabetes, 1 g/day of n-3 fatty acids produced serious vascular events in 8.9% against 9.2% on placebo (rate ratio 0.97, 95% CI 0.87 to 1.08, P=0.55) Bowman 2018, while 4 g/day of purified EPA in a different population did reduce them Bhatt 2019. Same nutrient class, four times the dose, a different molecule and a different answer.

Lp(a) is the number nobody on this page has measured and nothing on this page will change. It is set by the LPA gene, it is a substantial independent risk factor, and no diet, statin or supplement meaningfully lowers it. Measuring it once changes how aggressively everything else on the list is treated, which is the only reason to measure something you cannot move.

The structural failure of this whole goal is confusing a biomarker with an outcome. A compound that improves a marker of aging in a mouse, or a methylation clock in a spreadsheet, has not been shown to extend anything. The four variables above are on this page precisely because their end points were death and heart attacks rather than a score, and that is an uncomfortable comparison for most of the rest of the longevity goal.

And if the reason you came here was the interesting stuff, that is a different page and it should be read as one. Nutrient sensing — mTOR, AMPK & caloric restriction mimetics and Cellular senescence & senolytics are mechanistically serious and evidentially early, and reading them next to this one is the comparison this pathway exists to force. The muscle half is built on Substrate, cell volume & training capacity, and the particle half is ApoB & LDL particle reduction.

How you would know it was working, on a real read-out and a real timescale

Everything here is falsifiable on a number, which is the point of the page. The prediction: ApoB and blood pressure respond within weeks and predict events; the supplements at the bottom of the list will not move either one, and that non-movement is the honest result rather than a failed test.

• ApoB (Apolipoprotein B) at 6 to 8 weeks after any change. Hepatic LDL receptor upregulation reaches a new steady state within about two weeks of a dose change, and the remaining time is assay and biological variation. ApoB is preferred over LDL cholesterol here because a person with small dense particles carries more particles at the same cholesterol mass, and the particle count is what the causal evidence is about Baigent 2010.
• Lipoprotein(a) — Lp(a) once, ever. Concentration is roughly 90% heritable and stable across adult life, so a repeat is a repeat of the same fact. It exists on this list to change the threshold at which everything else gets treated.
• HbA1c (Hemoglobin A1c) at 12 weeks with Fasting Insulin beside it. Insulin moves years before glucose does, because compensation precedes failure, so a rising insulin with a flat HbA1c is the early finding and the one worth acting on.
• Comprehensive Metabolic Panel (CMP) at 12 weeks on a statin, and Cystatin C with eGFR if creatinine is confusing. Transaminase rises above three times the upper limit are uncommon and are the specific thing being watched for; creatine supplementation raises creatinine without changing filtration, so cystatin C is how those two get separated on a page that recommends both.
• Grip strength annually, on a dynamometer, same hand. It is the cheapest longevity measurement in existence, it tracks the variable with the largest observational signal on this page Celis-Morales 2018, and unlike the blood work it is a measure of something you built rather than something you took.

What will fool you. A single clinic blood pressure overstates by enough to change a treatment decision; the trial evidence rests on standardized, averaged measurements Wright 2015, and a home average over a week is closer to that than a rushed reading is. HbA1c is falsely low in anything that shortens red cell lifespan and falsely high in iron deficiency, so it should be read next to the Complete Blood Count (CBC) with Differential. And improvements that arrive in the first month of any of this are usually the other things that changed in the same month.

Sources read for these sections

• Baigent C. Efficacy and safety of more intensive lowering of LDL cholesterol: a meta-analysis of data from 170,000 participants in 26 randomised trials. The Lancet 2010;376(9753):1670-81 · PMID 21067804
• Wright JT. A Randomized Trial of Intensive versus Standard Blood-Pressure Control. New England Journal of Medicine 2015;373(22):2103-16 · PMID 26551272
• Celis-Morales CA. Associations of grip strength with cardiovascular, respiratory, and cancer outcomes and all cause mortality: prospective cohort study of half a million UK Biobank participants. BMJ 2018;361:k1651 · PMID 29739772
• Bowman L. Effects of n-3 Fatty Acid Supplements in Diabetes Mellitus. New England Journal of Medicine 2018;379(16):1540-1550 · PMID 30146932
• Bhatt DL, et al. Cardiovascular Risk Reduction with Icosapent Ethyl for Hypertriglyceridemia. The New England Journal of Medicine, 2019 · PMID 30415628
• Arem H. Leisure time physical activity and mortality: a detailed pooled analysis of the dose-response relationship. JAMA Internal Medicine 2015;175(6):959-67 · PMID 25844730
• Morton RW. A systematic review, meta-analysis and meta-regression of the effect of protein supplementation on resistance training-induced gains in muscle mass and strength in healthy adults. British Journal of Sports Medicine 2018;52(6):376-384 · PMID 28698222