The Evidence Library

The Evidence Library /

Every dose we print traces back to a published study on a specific ingredient at that dose. This page is the map: what was studied, at what dose, and what's actually in Small Hours. Every reference links to PubMed — check us.

KSM-66® Ashwagandha — 600 mg

Root-only extract, standardized to at least 5% withanolides — the extract used in the 2019 ashwagandha sleep trial listed below.

Daily doses used in research	250–600 mg
In Small Hours	600 mg a day, in one evening serving

• Chandrasekhar et al., 2012 — randomized, double-blind, placebo-controlled; 64 adults, 60 days, 600 mg/day KSM-66, given as 300 mg twice daily; measured perceived stress and serum cortisol. PubMed 23439798
• Langade et al., 2019 — randomized, double-blind, placebo-controlled; 60 adults enrolled as patients with a clinical sleep complaint and anxiety symptoms; 600 mg/day given as 300 mg twice daily, 10 weeks; measured time to fall asleep, sleep efficiency by wrist actigraphy, and sleep quality. PubMed 31728244
• Langade et al., 2021 — randomized, double-blind, placebo-controlled; 80 adults: 40 healthy volunteers and 40 patients with a clinical sleep complaint; 8 weeks; measured time to fall asleep, total sleep time, wake time after sleep onset and sleep efficiency. The authors report larger changes in the patient group. PubMed 32818573
• Salve et al., 2019 — randomized, double-blind, placebo-controlled; compared 250 mg and 600 mg daily for stress and sleep measures. PubMed 32021735

Who was in these trials: all participants in the 2019 trial, and half in the 2021 trial, were patients with a clinical sleep complaint. We don't claim Small Hours does what was measured there. Our label claim is narrower — supports restful sleep† — and rests on the saffron and L-theanine trials listed below. If you're being treated for a sleep condition, talk to your doctor before using any supplement.

Why root-only matters: the traditional material and the KSM-66 studies above use the root. Some competitors use leaf or leaf-heavy extracts to show bigger withanolide percentages — a different material than what was studied. We wrote a full breakdown.

L-Theanine — 200 mg

The amino acid found in green tea.

Daily doses used in research	200 mg (most common single dose)
In Small Hours	200 mg — the studied dose

• Hidese et al., 2019 — randomized, placebo-controlled; healthy adults, 4 weeks, 200 mg/day; measured stress-related symptoms and sleep quality. PubMed 31623400
• Rao et al., 2015 — a review of L-theanine research. PubMed 25759004

affron® Saffron Extract — 28 mg

Standardized extract of the stigma of Crocus sativus. Supports sleep quality over weeks.†

Dose used in the sleep research	28 mg/day, 1 hour before bed
In Small Hours	28 mg — the same extract, the same dose

• Lopresti et al., 2021 (Sleep Medicine) — 28-day, 3-arm, double-blind, placebo-controlled trial in 120 adults with unsatisfactory sleep; placebo vs 14 mg vs 28 mg affron® taken 1 hour before bed. Saffron improved sleep-quality ratings more than placebo. PubMed 34438361
• Lopresti et al., 2020 — randomized, double-blind, placebo-controlled trial of affron® (14 mg twice daily) over 28 days in 63 healthy adults with self-reported sleep problems. PubMed 32056539

Why the branded extract: the published sleep trials were run on affron® specifically, standardized to its own markers. 28 mg of an unnamed saffron extract is not the same dose — it’s a different material at the same weight.

What's deliberately not in the formula

Three ingredients, and every one of them is here because we can point to the dose it was studied at. No ingredient in this formula is there to look good on a label. No melatonin. No proprietary blends — every dose above is on the label. Other ingredients: hypromellose (vegetable capsule), rice flour, rice hull concentrate — the full Supplement Facts panel is on the product page. Why blends hide doses.

---

WE CHECKED THE SHELF / AUGUST 2026

The problem isn’t that they give you too little. It’s that you can’t tell.

We assumed most ashwagandha products were underdosed. So we checked, using the government’s own data — and we were wrong.

We pulled 336 ashwagandha product labels from the Dietary Supplement Label Database, run by the Office of Dietary Supplements at the National Institutes of Health, and read the numbers ourselves. Only 15% of the products that state a dose fall below 300 mg, the low end of the clinical range. Nearly half sit at 600 mg or above. On raw milligrams, the shelf is not as bad as we expected.

The real problem showed up somewhere else.

Of 336 labels	Share
State how standardized the extract is	29%
Do not state it at all — their dose cannot be compared to any study	71%
Of those that do state it: weaker than 5% withanolides	47%
Do not say which part of the plant was used	27%
Name KSM-66 specifically	18%

Here is why that matters more than the milligram count. Ashwagandha extract can be standardized to almost anything — in our sample the disclosed figures ran from 0.5% to 35% withanolides. Six hundred milligrams at 1% and six hundred milligrams at 5% are not the same product, and they are not off by a little. They are off by five times.

So when a label says “600 mg ashwagandha” and stops there, the number on the front is unverifiable. Not wrong — unverifiable. Seven labels in ten are in that position, and nothing about it is illegal.

The clinical trials behind ashwagandha used root extract standardized to 5% withanolides. That is the reason our label says all three things — the part of the plant, the standardization, and the dose. Any one of them alone tells you nothing.

We publish the uncomfortable half of this too: by raw dose, most of the market is fine, and our first assumption was wrong. We would rather correct ourselves in public than keep a claim that our own data does not support.

---

WHAT WE LEFT OUT

Thirteen ingredients didn’t make it into this bottle.

Every sleep brand publishes what it put in. Almost nobody publishes what they considered and rejected — which is strange, because that list tells you far more about how a formula was built.

Here is ours, with the actual reason in each case. Some were rejected for weak evidence. Some had fine evidence and simply would not fit. One of them we still think is interesting.

Considered	Why it isn’t here
Glycine, 3 g	The evidence is decent. Three grams is three or four extra capsules on its own. It belongs in a powder, not in this format.
Apigenin, 50 mg	Zero human trials of isolated apigenin for sleep. All the human data is on chamomile extract. It became popular through a podcast, not through research — and adding it would contradict everything else on this page.
Magnesium L-threonate	One positive trial run close to the vendor, one largely null trial in 2026. Expensive, bulky, and it delivers less elemental magnesium than a fully-reacted bisglycinate.
Lavender extract	Genuinely strong data, but on daytime mood, not sleep — sleep was only a secondary outcome there. It also needs a softgel we can’t combine with capsules, and it causes lavender burps. Interesting. Impractical.
Lemon balm	Best available data is open-label, with no placebo group, or buried inside combinations. Not the kind of proof we’re willing to print a dose against.
Valerian	Decades of inconsistent results and wildly variable extract quality. It also smells terrible, which matters more than it should for something you take every night.
Hops	Only ever tested in combinations. No way to know what the hops did.
Tart cherry	Two small pilot studies, both on juice. Juice is not a capsule, and we won’t pretend the result transfers.
GABA	Very limited evidence, and it’s still unsettled whether oral GABA reaches the brain at all.
5-HTP	No solid sleep trials, and it interacts with SSRIs. That alone rules it out.
Passionflower	The evidence is thin.
Melatonin, 0.3 mg the uncomfortable one	Here’s the irony we’re not going to hide. At a true physiological dose — 0.3 mg, roughly a tenth of what’s sold — melatonin has trial data showing it works mid-night. We still said no. Not because the research is bad, but because the shelf sells 3 to 10 mg, that’s what leaves people foggy at 7 a.m., and that’s the exact experience the people we make this for are trying to get away from. A melatonin line on our label would undo the reason they came. If you want to try 0.3 mg on its own, we think that’s a reasonable thing to do. It costs a few dollars and it isn’t ours.

Why there is no magnesium. We looked hard at magnesium bisglycinate and left it out. The evidence is thinner than its reputation on the sleep aisle: one independent trial of this exact form (250 mg, 155 adults, four weeks) showed a modest improvement on a sleep-quality score — effect size 0.2 — with every secondary outcome null (PubMed 40918053), and the systematic review of older trials calls the literature too weak for firm recommendations (PubMed 33865376). None of it measured time awake after sleep onset — the 3 a.m. problem. Two other things tipped it. At a meaningful elemental dose, a fully-reacted chelate weighs about 2 grams per serving — three capsules on its own — for an ingredient with no measured effect on night waking. And across several independent sources, a recurring minority report that glycinate forms taken at bedtime wake them in the night; anecdote, not trial, but in a formula for people who wake at 3 a.m., we would rather leave the variable out than argue with it. If you want magnesium, take it: 200–250 mg elemental of a well-absorbed form, in the morning, is a reasonable and cheap experiment — and it isn’t ours.

Three ingredients made it through. The other thirteen are on this page because a formula is defined as much by what you refuse to add as by what you include — and because the easiest way to make a supplement look impressive is to add things nobody will check.

Added 16 September 2026. Phosphatidylserine was not among the ingredients we considered at formulation — it reached us later, as the internet’s favorite answer to 3 a.m. waking, so we counted it the way we count everything. PubMed holds zero randomized trials of phosphatidylserine for sleep in adults; its cortisol effect was shown in two daytime laboratory stress-test trials of a PS–phosphatidic acid complex (PAS) at 400 mg, not PS alone, while 57% of dedicated PS labels sell 100 mg and 71% sell it as a memory supplement. Not a verdict on the molecule — a verdict on the evidence. The full count is in the Journal →

---

This page is educational, summarizes research on individual ingredients — not on Small Hours as a finished product — and is not medical advice. Study outcomes reflect the specific populations and conditions studied; individual results vary.