Issue #20 — The Dose Nobody Told You About: What the Lancet's ADHD Sweet Spot Study Means for Founders
🎯 TL;DR
The most comprehensive ADHD dosage study ever conducted just mapped the exact point where every major stimulant stops working better and starts working worse. If you've been quietly creeping your dose up because focus feels "off," the data says you might be chasing the wrong variable.
This week: How to use the new Lancet meta-analysis to run a performance audit on your meds — without firing your psychiatrist.
Read time: 6 minutes
💊 The Dose You Quietly Increased Last Quarter
You know the feeling. The 30mg Vyvanse used to clear the fog by 9am. Then it didn't, so your prescriber bumped you to 40mg. Then 50mg. Now you're sweating through investor calls, your resting heart rate looks like a cardio session, and you're still rage-rewriting the same pitch deck slide at 11pm. You assumed this was the cost of doing business. The Lancet just said: probably not.
On 14 May 2026, Lancet Psychiatry published the most comprehensive dose-effect network meta-analysis of ADHD medications to date, led by a team at the University of Southampton and pooling decades of randomized trial data across five drugs: methylphenidate, amphetamines, atomoxetine, guanfacine, and viloxazine. The headline finding, in the authors' own words: there is "little gain above optimal limits." Past a certain dose, efficacy flatlines. Side effects don't.
🔬 What the Southampton Team Actually Found
The analysis modeled dose-response curves separately for symptom reduction and tolerability — meaning they didn't just ask "does it work?" They asked "at what point does the cost outweigh the benefit?" Across adults, every stimulant traced the same shape: a steep climb in efficacy at low-to-moderate doses, then a plateau. Tolerability, meanwhile, kept degrading linearly. The optimal zone — best symptom control per unit of side effect — landed lower than the doses many adult founders are actually prescribed.
A few specifics worth knowing before your next appointment:
| Medication | Adult "sweet spot" zone | What happens above it |
|---|---|---|
| Methylphenidate | ~30 mg/day equivalent | Efficacy plateaus; insomnia, appetite loss, BP climb |
| Amphetamines | ~20–30 mg/day equivalent | Diminishing returns; anxiety and cardiac load rise |
| Atomoxetine | ~80 mg/day | Marginal symptom gain; GI and sleep cost increase |
| Guanfacine | ~3–4 mg/day | Sedation outpaces focus benefit |
| Viloxazine | ~200–400 mg/day | Limited dose-dependent efficacy gain |
(These are population-level optimums, not your personal prescription — but they're the goalposts your psychiatrist is now working against.)
"The data suggest clinicians have been operating with more dose latitude than the evidence supports — and patients have been absorbing the side-effect cost of that uncertainty." — Lancet Psychiatry, dose-effect meta-analysis, May 2026
The new paper builds on earlier work by Samuele Cortese and colleagues, whose 2018 Lancet Psychiatry network meta-analysis ranked ADHD drugs by overall efficacy. The 2026 paper adds the axis that's been missing for adult prescribing: how much of each, not just which. For founders self-tracking with Apple Watch, Whoop, or even just a sleep log, that's now a measurable target instead of a vibe.
🛠 The Prescriber Conversation As Performance Audit
Walk into the appointment the way you'd walk into a board meeting about a marketing channel that's underperforming. You're not asking permission. You're presenting data and proposing a test.
- Bring two weeks of structured data. Daily dose, time taken, subjective focus score (1–10) at 10am, 1pm, and 4pm, resting heart rate (your wearable already tracks this), sleep onset time, and appetite. A simple Notion table works. Without this you're arguing from feelings. With it you're arguing from a dataset, which most psychiatrists will respect immediately.
- Name the Lancet finding directly. Try: "There's a May 2026 Lancet Psychiatry network meta-analysis showing efficacy plateaus around [X] mg for [your drug]. I'd like to see whether I'm sitting above my personal optimum." This signals you're an informed partner, not a dose-shopper — an important distinction in a chart note.
- Propose a structured down-titration, not a cut. "Could we step down 10mg for four weeks while I keep tracking, then re-evaluate?" Frame it as an A/B test with a rollback plan. Clinicians say yes to experiments far more readily than to "I want less."
- Pre-define the failure condition. Decide in advance what data would tell you the lower dose isn't working — e.g., focus scores drop more than 2 points for five-plus days in a row. This protects you from the "I feel weird, abort" reaction on day three, which is almost always adjustment noise, not true ineffectiveness.
If your prescriber refuses to engage with the literature at all, that itself is data. Get a second opinion. Psychiatrists who take adult ADHD seriously are reading this paper this month.
⚡ The ADHD Angle
Here's the uncomfortable part. ADHD brains are dopamine-seeking by architecture — and a dose increase feels like progress in a way that almost nothing else in your day does. You take the higher pill. Within an hour, something is different. Your brain logs "more = better" and the loop closes. That's not focus returning; that's noradrenaline cranking. The Lancet curve is essentially the empirical shape of that illusion.
Layer in RSD and founder identity, and it gets worse. When focus dips, you don't think "my dose might be slightly off the sweet spot." You think "I'm failing, I need to fix me, fix me harder." The dose creep is a self-worth response disguised as a clinical decision. Executive function research suggests the prefrontal regions stimulants act on are non-linear — too much agonism past a point actually degrades the signal-to-noise ratio you're trying to improve. Translation: the 50mg version of you may literally be making worse decisions than the 30mg version, even though it feels sharper. Hyperfocusing on the spreadsheet at the cost of missing what your co-founder actually said in the meeting is not a win.
🎯 This Week's Challenge
- Start the 14-day audit today. Spin up a Notion or Apple Notes table with five columns: dose, focus 1–10 (3x/day), resting HR, sleep onset, appetite. No analysis yet — just capture.
- Book the appointment for week three. Don't wait for your next routine refill. Email your prescriber now, mention you've been reading the May 2026 Lancet Psychiatry dose-effect meta-analysis, and request 20 minutes to review your data.
- Pick one variable to defend first. Sleep onset is the cheapest signal of overshoot. If you're falling asleep after 1am on workdays, that's the first number to bring up.
See you Tuesday, L-P
P.S. — I'm not your doctor and this isn't medical advice. But you're allowed to read primary research, and you're allowed to bring it into the room. Forward this to the founder friend who keeps complaining their meds "stopped working." Maybe they didn't.
Divergent — Strategy for brains that don't do boring.