My Current Supplement Stack — and the 7 I Tried and Quit
By Akash S. Chauhan | First Principles Healthspan, Issue 04
The supplement industry runs on optimism and ambiguity. Mechanistically plausible. Studied in rodents or in vitro. One small pilot trial that never replicated. Priced accordingly. I have bought into this cycle more times than I care to admit, so this issue is a post-mortem as much as a recommendation list.
Why I'm writing this
I get asked about my supplement stack regularly. The honest answer is not a clean list of ten things to buy. It is a smaller list of things that survived rigorous self-application of one criterion: does the human trial evidence justify the cost, the hassle, and the potential downside? The quit list is longer than the keep list, and the quit list is the more useful half.
Nothing in this issue is medical advice. If you have a chronic condition, are on medications, or are pregnant, talk to a clinician before adding anything. This is what I do, why I do it, and what I stopped doing.
What I actually take (and why each earns its place)
1. Creatine Monohydrate
Dose: 5g/day, no loading phase, any time of day, with or without food.
Evidence: Creatine monohydrate is one of the most studied compounds in sports nutrition. The meta-analysis by Rawson and Volek (J Strength Cond Res, 2003) across 22 studies found consistent improvements in high-intensity strength and power output. Beyond muscle, the picture is getting more interesting: Rae et al. (Proc Royal Soc B, 2003; PMID: 14561278) ran a double-blind RCT in vegetarians and found significant improvements in working memory and fluid intelligence with 5g/day creatine supplementation. The proposed mechanism is ATP resynthesis in the brain, which faces similar energy demands to skeletal muscle.
Why it made the cut: Safety record is unambiguous across decades of data. Cost is negligible — roughly $0.10/day for monohydrate. The strength and power evidence is as clean as sports science gets, and the cognitive angle, while not yet definitive, is worth the essentially-zero downside at this dose.
Where to get it: Thorne Creatine is what I use — third-party tested, no filler. Any pharmaceutical-grade monohydrate works. Do not pay a premium for "Kre-Alkalyn" or other branded forms; the bioavailability evidence favors plain monohydrate.
2. Omega-3 (EPA + DHA)
Dose: 2–3g combined EPA+DHA per day from a high-quality fish oil or algae-based supplement.
Evidence: The REDUCE-IT trial (Bhatt et al., NEJM 2019; PMID: 30415628) showed icosapentaenoic acid (EPA) at 4g/day reduced major adverse cardiovascular events by 25% in patients with elevated triglycerides already on statins. That is a meaningful effect size in a high-risk population. The caveat everyone should know: the placebo arm used mineral oil, which may have inflated LDL-C and hsCRP in controls — making EPA look better than it might in a true placebo comparison. The ORIGIN trial (NEJM 2012) found no benefit from 1g/day EPA+DHA in a different patient population. The dose and patient context matter enormously.
At the population level, habitual fish consumption associates with lower cardiovascular mortality in prospective cohort data, but supplementation trials at moderate doses (1g/day) have been inconsistent. The honest read: high-dose EPA has the strongest cardiovascular signal, but I am not on a statin and I am not in the REDUCE-IT risk profile. My rationale at 2–3g/day sits more on the anti-inflammatory and triglyceride-reduction data, both of which have solid dose-response evidence.
Why it made the cut: Anti-inflammatory effect is real at doses above 2g EPA+DHA. Triglyceride reduction is well-established (roughly 15–30% at 3–4g/day). I eat oily fish 2–3x/week but supplementing bridges the gap. I get my levels tested annually (HS-Omega-3 Index; target >8%).
Where to get it: Momentous Omega-3 — high EPA:DHA ratio, third-party tested for oxidation and heavy metals. Oxidized fish oil may be worse than no fish oil; buy from brands that test TOTOX values.
3. Magnesium Glycinate (or Threonate)
Dose: 300–400mg elemental magnesium per day, taken in the evening.
Evidence: NHANES data consistently shows roughly 45–50% of Americans fail to meet the estimated average requirement for magnesium from diet alone. The mineral is a cofactor in over 300 enzymatic reactions, including ATP synthesis, DNA repair, and protein synthesis. For the specific claim that draws me to it: Abbasi et al. (J Res Med Sci, 2012) ran a randomized, double-blind, placebo-controlled trial in elderly adults with insomnia and found that 500mg magnesium per day improved sleep time, sleep efficiency, serum melatonin, and serum cortisol relative to placebo. Effect sizes were modest but directionally consistent.
Why it made the cut: The risk-adjusted calculus here is favorable. Magnesium from food is broadly safe. Supplemental magnesium at these doses is well-tolerated in the glycinate or threonate forms. The oxide form has poor bioavailability (~4%) and is mostly what cheap supplements use — pay attention to the form. Given my diet skews lower in magnesium-dense foods (dark leafy greens, legumes, nuts) some weeks, this fills a gap I can verify with periodic serum testing (though red blood cell magnesium is more accurate than serum).
4. Vitamin D3 + K2
Dose: 2,000–4,000 IU D3 paired with 100mcg MK-7 K2, taken with a fat-containing meal.
Evidence: Context-dependent. I take this only because my 25-hydroxyvitamin D level was 28 ng/mL before supplementation — below the 40 ng/mL floor I target. The VITAL trial (Manson et al., NEJM 2019; PMID: 30415629) randomized 25,871 adults to 2,000 IU D3/day or placebo and found no reduction in the primary composite cardiovascular endpoint, but did find a statistically significant reduction in cancer mortality (24% lower in the vitamin D group) — notably in people with normal BMI. That is a secondary endpoint finding, which tempers how strongly I weight it, but it is directionally interesting.
The K2 pairing is about directing calcium metabolism — the theoretical concern is that D3 supplementation increases intestinal calcium absorption, and K2 (specifically MK-7) activates matrix Gla-protein and osteocalcin, which in theory routes calcium toward bone and away from arterial walls. The human outcome data for this specific combination is less robust than I would like; it is mechanistically sound but not trial-proven at this pairing.
Why it made the cut: Blood level optimization, not population-level supplementation dogma. If your 25(OH)D is already above 50 ng/mL, you probably do not need this. Sun exposure is the better source when practical. I test quarterly and adjust the dose accordingly.
5. Protein Powder (Whey or Collagen as needed)
Dose: Variable; used as a tool to hit 1.6g of protein per kilogram of bodyweight per day.
This is not a supplement in the pharmacological sense — it is food in a more convenient form. The target of 1.6g/kg/day is derived from the meta-analysis by Morton et al. (Br J Sports Med, 2018) across 49 RCTs, which found that protein intakes beyond that threshold produced no additional benefit for resistance-trained adults. I use whey or a blend on training days when whole-food protein is harder to fit. No mystery to it. Protein powder on Amazon.
The 7 I Tried and Quit
1. NMN / NR (NAD+ Precursors)
What I expected: Meaningful restoration of declining NAD+ levels. The Sinclair lab work on sirtuins and NAD+ biology is compelling, and the animal model data is striking.
What happened: I ran NMN at 500mg/day for four months. Subjectively: nothing I could attribute to it that I could not also attribute to better sleep that period. No measurable performance changes on tracked workouts.
What the evidence actually says: Yoshino et al. (Cell Metabolism, 2021) ran a placebo-controlled trial of NMN in postmenopausal women with prediabetes and found improved insulin sensitivity in skeletal muscle. That is one positive trial in a specific population. Other trials in healthy adults have found NAD+ levels do rise with supplementation, but functional outcomes have been inconsistent or null. The dose required to produce meaningful systemic NAD+ elevation in humans appears to be higher than the 250–500mg most people take, and the cost at 1,000mg+/day is substantial. Quit reason: cost-to-evidence ratio is poor at current prices. I will revisit as trial data matures and costs drop.
2. Berberine
What I expected: "Natural metformin" — the framing everywhere in the longevity-adjacent community. A Dong et al. meta-analysis (Evidence-Based Medicine, 2012) covering 14 RCTs found berberine reduced HbA1c and fasting glucose comparably to several oral hypoglycemics in type 2 diabetic patients.
What happened: GI side effects that ranged from inconvenient to disruptive. Bloating and loose stools that did not resolve after the first two weeks.
What the evidence actually says: The glucose management data exists, but it is largely in people with frank metabolic dysfunction, not in healthy adults self-hacking. Bioavailability is poor and variable. Berberine inhibits CYP3A4, which creates real drug interaction risk if you are on anything metabolized by that pathway. Self-prescribing a compound with these properties to address glucose levels that are already in a healthy range is a risk-reward calculation that does not hold up. Quit reason: the GI burden, the interaction profile, and the mismatch between the target population in the evidence and my situation.
3. Collagen Peptides (Beyond Protein Contribution)
What I expected: Joint-specific benefits — reduced pain, improved cartilage matrix. The marketing is detailed and sounds mechanistic.
What happened: No subjective change in joint comfort over a three-month trial. The product I was using was also significantly more expensive per gram of protein than whey.
What the evidence actually says: The amino acid profile of hydrolyzed collagen is heavily skewed toward glycine and proline, which are conditionally essential but not limiting in most diets. The hypothesis that orally ingested collagen peptides are preferentially incorporated into joint cartilage has weak direct human trial support. Shaw et al. (Am J Clin Nutr, 2017) found co-ingesting gelatin with vitamin C and preceding exercise increased collagen synthesis markers in connective tissue — interesting, but a far cry from clinical joint-health outcomes. Quit reason: the premium cost is not justified when regular dietary protein achieves a comparable amino acid intake. If joint health is the goal, resistance training and managing body weight have better evidence.
4. Ashwagandha
What I expected: Measurable cortisol reduction and stress buffering. Chandrasekhar et al. (Indian J Psychol Med, 2012) ran a double-blind RCT finding significant improvements in stress, anxiety, and cortisol in chronically stressed adults at 300mg root extract twice daily.
What happened: Modest, inconsistent. I tracked HRV throughout. Trends were not outside normal variation. Effect, if present, was below my detection threshold.
What the evidence actually says: The stress and cortisol data is real but heterogeneous across studies — effect sizes vary widely depending on the extract, dose, and baseline stress level of the population. Long-term safety data beyond 3 months is thin. There are case reports of hepatotoxicity with high doses, though causality is not established. For someone with high baseline stress, this may earn a place. For me, sleep hygiene and training load management moved the HRV needle more reliably. Quit reason: insufficient signal above noise for my use case, and I am not comfortable with the long-term safety unknowns at sustained use.
5. Resveratrol
What I expected: SIRT1 activation, improved mitochondrial function. Again, Sinclair lab. The mechanistic story is elegant.
What happened: Nothing I could detect.
What the evidence actually says: Resveratrol's oral bioavailability is poor — it is rapidly metabolized in the gut and liver, and peak plasma concentrations from typical supplement doses are low. Human RCT data has been consistently disappointing relative to the animal work. Specifically, high-dose resveratrol has been found in some studies to blunt the adaptive signaling response to exercise (Gliemann et al., J Physiol, 2013 — reduced training-induced improvements in VO2max in older men). That is the opposite of what I want. The "absorbed better from red wine" line is tongue-in-cheek, but the practical truth is that the amounts in wine are pharmacologically trivial and the wine itself has costs. Quit reason: poor bioavailability, disappointing human trial data, and potential interference with exercise adaptation.
6. Pre-Workout Stimulants
What I expected: Better performance in morning training sessions. Caffeine + beta-alanine + whatever else was in the formula.
What happened: Effective acutely, tolerance within 3–4 weeks. Started needing it to feel normal training sessions rather than getting a boost above baseline. More critically: when I tracked HRV and resting heart rate, any use within 6 hours of my target sleep time reliably degraded HRV by 8–12% and increased resting HR the following morning.
What the evidence actually says: Caffeine's ergogenic effects are well-established and not in question. The issue is that its half-life of 5–7 hours means a 200mg dose at 4pm still has 100mg active at 9pm in most people. The sleep disruption compounds — poor sleep reduces the recovery that training is supposed to drive. The net effect over weeks was negative. Quit reason: the acute performance gain does not survive accounting for the downstream sleep and recovery cost. I use 100–200mg caffeine before training now, only before morning sessions.
7. Melatonin at 5–10mg
What I expected: Better sleep onset, maybe some additional antioxidant effects from high doses.
What happened: Groggy mornings. The "melatonin hangover" is real at 5–10mg and is physiologically predictable — supra-physiological doses have a longer effect tail.
What the evidence actually says: Physiological melatonin secretion peaks at roughly 0.1–0.3mg in the bloodstream. Most supplements sell 5–10mg, which is 20 to 100 times the physiological dose. The Circadin data and most sleep-shifting literature uses 0.5mg, taken 1 hour before the target sleep time, to shift circadian phase. At high doses, there is concern about feedback suppression of endogenous melatonin synthesis with chronic use, though robust long-term human data is limited. Quit reason: replaced the high-dose habit with 0.5mg when needed for travel or circadian shifting, and invested in the higher-leverage interventions: consistent wake time, blocking short-wavelength light in the evening, and cool room temperature. Sleep hygiene moved the needle more than any dose of melatonin.
This Week's One Thing to Do
If you take nothing else, consider creatine monohydrate. It has one of the strongest combined safety and efficacy profiles of any compound in the supplementation literature — decades of data, no credible safety signal at 3–5g/day, consistent positive effects on strength and power output, and emerging (if not yet definitive) cognitive data. Dose: 3–5g/day. Time of day does not matter. With food or without. No loading phase required. It is roughly $20 for a three-month supply in monohydrate form. The cost-to-evidence ratio is better than almost anything else on the shelf.
Sign-off
The supplement industry defaults to optimism because optimism sells. Running a personal trial with tracked outcomes and a willingness to quit is how you separate signal from marketing. Most of the quit list above has some evidence behind it — that is what made it appealing enough to try. The issue is that evidence quality, dose-response specificity, and individual context determine whether a mechanism becomes a result.
The keep list is short. That is the point.
If you want me to go deeper on any of the above — or have a supplement you want me to evaluate using the same framework — reply to this email. I read everything.
Until next week, Akash S. Chauhan
Education only. Not medical advice. Always consult a licensed clinician for individual decisions.