Attention Deficit (Hyperactivity) Disorder

1 The case for treating at all

The first decision is whether to medicate. The vitalist / anti-pharma instinct — that medicating a personality trait is somehow a capitulation to a productivity-obsessed society — deserves a serious hearing, but it needs to contend with the actuarial data.

Untreated ADHD in adults is associated with reduced life expectancy — Barkley’s estimates run to ~13 years, driven by accidents, substance use, suicide risk, and the accumulated health consequences of impulsivity (Barkley and Fischer 2022). This seems wildly too big, so let us forget the number and just say “big difference”. Indeed, everyone agrees it is big: 2024 umbrella review of the impacts of ADHD found roughly doubled risk of injuries, significantly elevated mortality, and pervasive functional impairment across education, employment, and relationships (Faraone et al. 2024). A population-based cohort study found that episodes of ADHD medication use were associated with reduced all-cause mortality and unintentional injuries, with stimulants specifically showing lower mortality risk (Li et al. 2024).

The relevant comparison, then, is not “medication versus a hypothetical neurotypical baseline” but “medication versus the counterfactual of an unmedicated idiosyncratic brain navigating traffic, relationships, tax returns, and the impulse-purchase aisle”. If the intervention is preventing car accidents and early death, the bar for tolerating side effects should be permissive.

That said, the political-economy critique isn’t vacuous. There is something uncomfortable about a society that pathologises attention variance while simultaneously engineering maximal distraction (Cassidy 2015; Pedersen, Albris, and Seaver 2021; Doran 2017). In a political economy dominated by the war for salience, what niche does ADHD fill? Is ADHD a risk factor for exploitation by the attention economy, an adaptive response to it, or a side effect of measuring productivity in ways that don’t suit every cognitive architecture? The answer is probably “yes” to all three.

2 Decision framework

The currency of decision in ADHD treatment should be something like quality-adjusted functioning — the difference between a day where you can sustain attention on the task that matters and a day where you can’t.

For each candidate:

1. Efficacy. Effect size on core symptoms (inattention, hyperactivity, impulsivity), ideally from meta-analyses. ADHD pharmacotherapy has unusually large effect sizes for psychiatry — SMDs of 0.4–0.7 — so the “does it work at all” question is not too troubling.
2. Tolerability and side effects. The common ones (appetite suppression, insomnia, cardiovascular effects) and the tail risks.
3. Tolerance and habituation. Does the effect wear off? Do we need escalating doses?
4. Substance-abuse risk. Given ADHD’s baseline elevated risk of substance-use disorders, does the medication make this better or worse?
5. What it precludes or enables. Sleep quality, exercise capacity, appetite, social functioning.

The distinguishing feature of ADHD pharmacotherapy compared to nootropics is that the untreated state is itself hazardous compared to some “average” baseline — “you are meaningfully more likely to crash your car or lose your job”. The default is probably “treat unless you have a specific reason not to”, because the costs of non-treatment are large and well-documented. (More on what it means that these are the same drugs as the nootropics in the medicalisation section below.)

3 Stimulants: methylphenidate and amphetamines

First-line for a reason. The Lancet Psychiatry’s network meta-analysis — 133 RCTs — found amphetamines the most efficacious class for adult ADHD, methylphenidate close behind, both clearly superior to non-stimulants (Cortese et al. 2018). A 2024 update confirmed stimulants and atomoxetine are the only pharmacological interventions with clear short-term efficacy in adults (Defined and Defined 2024).

A caveat on the evidence base: ~41% of ADHD medication trials are industry-funded, and the vast majority carry high risk of bias on Cochrane assessment (Defined and Defined 2024; Punja et al. 2016). Stimulant trials are particularly vulnerable to functional unblinding — the physiological effects (appetite suppression, elevated heart rate) are noticeable enough that participants and clinicians can often guess who got the active drug, which inflates subjective outcome measures. When the 2024 Lancet review restricted to low-risk-of-bias trials, stimulants still beat placebo on self-reported symptoms, which is reassuring but not fully exonerating — “low risk of bias” on formal Cochrane criteria doesn’t guarantee successful blinding in a stimulant trial. The short version: stimulants almost certainly work, the effect sizes reported in meta-analyses are probably somewhat inflated, and the long-term evidence is thinner than the confident tone of most prescribing guidelines would suggest.

Methylphenidate (Ritalin, Concerta, etc.) blocks reuptake of dopamine and noradrenaline. Extended-release formulations are generally preferable — smoother pharmacokinetics, lower abuse potential, better mapping onto a working day. Moderate-to-large effect on core symptoms; also improves quality of life (Coghill et al. 2024). For the side-effect profile in detail see Kowalczyk et al. (2025).

Amphetamines (dextroamphetamine, lisdexamfetamine / Vyvanse, mixed salts / Adderall) block reuptake and promote release of dopamine and noradrenaline. Slightly larger effect sizes than methylphenidate, slightly more aggressive side effects — especially appetite suppression (~2 kg average weight loss over placebo). Lisdexamfetamine is a prodrug (inactive until enzymatically cleaved in blood), which makes it the stimulant with the lowest abuse potential and the smoothest absorption curve.

Both classes share the same side-effect family: appetite suppression, insomnia (timing-dependent), modest cardiovascular effects (~3–5 mmHg systolic, ~1–2 bpm), dry mouth, jitteriness; amphetamines sometimes add anxiety or irritability. See side effects and management below.

4 Non-stimulants

Atomoxetine (Strattera): selective noradrenaline reuptake inhibitor. Similar efficacy to stimulants in short-term RCTs but worse tolerability (higher dropout) and slow onset (4–6 weeks vs same-day) (Defined and Defined 2024). No abuse potential, no appetite suppression, 24-hour coverage, better for comorbid anxiety. Use for: stimulant non-responders (~20–30% of adults), comorbid anxiety or substance-use disorder, stimulant-induced insomnia. Rare tail risks: liver injury (very rare, boxed warning in some jurisdictions), suicidal ideation in young adults (small signal, similar to SSRI labelling).

Guanfacine / clonidine (α₂-agonists): modest effect sizes, better for hyperactivity/impulsivity than inattention. Guanfacine preferred — more selective, longer half-life, less sedation (Sallee et al. 2009). Main side effect is sedation (useful if insomnia is comorbid). Must taper — rebound hypertension on abrupt discontinuation. Useful as an adjunct to stimulants or when stimulant-induced insomnia is the dose-limiting side effect.

Bupropion (off-label): noradrenaline-dopamine reuptake inhibitor. Better than placebo but below stimulants and atomoxetine (Cortese et al. 2018). Worth knowing about when ADHD is comorbid with depression or smoking.

5 Side effects and their management

Cardiovascular: The most serious long-term concern. Short-term RCTs show “relatively small” increases in blood pressure and heart rate (Zhang et al. 2025). Longer-term observational data is mixed — one study found increased hypertension risk (Shin et al. 2024), another large registry meta-analysis found no association with cardiovascular events at the population level (Hesselbrock et al. 2024). Practical upshot: baseline ECG and blood pressure; monitor every six months; cardiology opinion if pre-existing risk. Worth weighing against the cardiovascular consequences of untreated ADHD.

Sleep: ADHD itself disrupts sleep, and stimulants can make it worse or better — a large clinical study found stable treatment associated with lower rates of insomnia than untreated ADHD (Hvolby 2015). When stimulants do cause insomnia: adjust timing, switch formulations, try melatonin for sleep-onset delay, bright light therapy in the morning, or guanfacine at bedtime (Wynchank, Bijlenga, and Kooij 2021).

Appetite: Stimulants suppress appetite reliably, more so with amphetamines. “Rebound hunger” at night can shift intake towards evening binge-eating. Management: substantial breakfast before peak effect, scheduled lunch even when not hungry, high-protein meals. Lisdexamfetamine has a separate indication for binge-eating disorder, so in the overweight patient this side effect might be a feature.

Substance abuse: The fear, and the evidence is reassuring. A meta-analysis found stimulant treatment neither increased nor decreased later substance-use risk (Humphreys, Eng, and Lee 2013); larger registry studies found a ~31% lower rate with treatment, and longer duration predicted lower risk (Z. Chang et al. 2014). Misuse is primarily associated with IR formulations, younger age, and concurrent substance-use disorders (Faraone et al. 2025) — prescribing ER stimulants to an adult with ADHD is not meaningfully riskier than prescribing an SSRI.

6 Tolerance and habituation

This deserves its own section because it is the single most common anxiety people have about starting stimulant medication, and the folk wisdom (“you’ll need more and more and eventually it won’t work”) is not well-supported by the clinical data.

Handelman and Sumiya (2022) reviewed the tolerance literature for ADHD stimulants and found: tolerance to subjective effects (euphoria, mood lift, “liking the drug”) develops within about two weeks of daily use, consistent with downregulation of reward circuits. Tolerance to therapeutic effects (attention, hyperactivity, impulsivity) at clinical doses was not clearly demonstrated in any of the studies they reviewed. The distinction matters enormously: the recreational user escalates doses chasing the euphoria; the clinical user doesn’t need the euphoria, and the effect they do need appears to be stable.

Tamminga et al. (2025)’s 2025 systematic review reached a similar conclusion from a larger evidence base: “preliminary evidence of tachyphylaxis to affective or behavioural effects; little or no evidence for tolerance to therapeutic or cardiovascular effects in the longer term.” They also note the field is under-studied — only 17 studies met their inclusion criteria — and that absence of evidence is not evidence of absence.

What “wearing off” means in practice: When patients report that their medication “stopped working”, the differential includes true pharmacological tolerance but also several confounds that are hard to disentangle: dose became suboptimal as circumstances changed (new job demands, worse sleep, more stress), comorbid conditions (depression, anxiety) became dominant, or the initial contrast effect faded — the first week on stimulants often feels dramatic, the new baseline becomes normal, and “normal” feels like “it stopped working” (Handelman and Sumiya 2022; Yanofski 2011). One long-term follow-up found only 2.7% of patients lost response without a clear external explanation; a separate study found ~25% reported tolerance within days to weeks (Handelman and Sumiya 2022). We don’t have clean data on the relative frequency of these explanations, and “perceived loss of efficacy” is probably a mix of all of them in proportions that vary by patient.

Drug holidays: Not well-supported for preventing tolerance in adults. The main evidence base is in children, where breaks are sometimes used to assess whether the medication is still needed or to allow catch-up growth. In adults, the consensus position is that drug holidays are not an optimal strategy and often result in return of symptoms and functional impairment (Ibrahim and Donyai 2015). If the question is “do I still need this?”, a supervised taper with objective symptom monitoring might be better than a self-directed holiday. jkjk no-one can afford to pay psychs to do that, so the next best thing is probably to quantify yourself, tracking the symptoms that matter before and after a break, and if you find the break is more disruptive than you expected, that’s good data to have for the next conversation with your doctor.

7 Adjuncts and supplements

Does anything stacks usefully on top of ADHD uppers?

L-Tyrosine comes up most often in ADHD forums — the logic being that stimulants burn through dopamine, tyrosine is a dopamine precursor, therefore supplementing tyrosine should help replenish the supply. The biochemistry is not wrong, but the clinical evidence is scarce. A 1987 open trial found initial improvement that disappeared within weeks (Reimherr et al. 1987). A 2015 review concluded tyrosine can help cognition under acute stress in neurotypicals, but only when dopamine stores are already depleted — and all studies were in non-ADHD populations (Jongkees et al. 2015). No RCT in ADHD populations exists.The theoretical story is plausible enough that I wouldn’t argue with someone who takes it, but the “stimulants deplete dopamine and tyrosine fixes that” narrative doesn’t seem complete: According to my friendly neighbour LLM, stimulants at clinical doses primarily block reuptake rather than deplete stores, and tyrosine hydroxylase (the rate-limiting enzyme) is feedback-regulated, so flooding the system with precursor doesn’t necessarily increase synthesis. File under: plausible mechanism, no clinical evidence, cheap enough that it’s probably doesn’t hurt to try.

Omega-3 fatty acids (EPA/DHA): the best-supported nutritional adjunct. Meta-analyses find a small but consistent effect on ADHD symptoms, well below stimulant effect sizes but potentially additive (J. P.-C. Chang et al. 2018). Low-certainty evidence overall; most useful as an adjunct rather than a replacement. Cheap, safe, probably worth taking if not already eating oily fish regularly — which is the same conclusion as on the ageing page, arrived at by a different route.

Zinc, magnesium, iron: each has some evidence of lower levels in ADHD populations (why?), and supplementation in deficient populations appears to help (Skalny et al. 2020). As with most vitamin and mineral supplements, “in deficient populations” is an operative phrase—blanket supplementation in non-deficient adults has weak support. Testing levels and correcting deficiencies is reasonable; a prophylactic “ADHD supp stack” is not well-evidenced.

Melatonin for stimulant-induced sleep-onset delay has genuinely good RCT support in the ADHD context — not for core symptoms, but for the most common side effect of the primary treatment (Wynchank, Bijlenga, and Kooij 2021).

8 What even is mindfulness?

I don’t know. See meditation for now.

9 ADHD medication as backdoor transhumanism

These are the same drugs that show up on the cognitive enhancement page. Methylphenidate and amphetamines don’t check diagnostic status before they work; the line between “treating a disorder” and “enhancing cognition” is drawn by a threshold, not by a different molecule. Sam Atis makes this point explicitly.

Gavin Leech calls this transnormalism: widespread cognitive enhancement happening through medicalisation rather than through any explicit transhumanist programme. Millions of people taking dopaminergic cognitive enhancers daily, socially unremarkable because it’s framed as treatment rather than augmentation. The medical label functions as a licensing mechanism — it gets insurance, clinical monitoring, titration expertise, and legal protection that the freelance biohacker buying grey-market research chemicals doesn’t get. I’d be more comfortable if classifying something as a disease were not a prerequisite for treating it, but here we are.

ADHD-as-disease is a pragmatic ‘fiction’: a lossy compression of continuous variation into a label, because the medical system needs labels to operate. Or, to flip this around into an IMO completely equivalent framing: Neurotypicality is also a fiction. People are wildly and bafflingly different.

If you are cynical you can claim it is Illich’s social iatrogenesis, but it looks like a mild and possibly benign form if so. IMO it doesn’t matter much whether ADHD is a “real” “disorder” in the sense of being a natural nosological kind. The suffering is real, the treatment helps, and whether we call it “treating a disorder” or “helping people thrive” is a question about which bureaucratic apparatus gets activated. It is lucky that this particular misfit to modernity happened to have a clean enough symptom profile and a good enough pharmacological target to mobilise the medical establishment; not every form of human difficulty gets that deal.

There are people near to me for whom it is very important that ADHD be “real”. Me complaining that it is not “real” is not a comment on the lived experience of people who find this diagnosis meaningful, so much as it is me complaining about the taxonomy. You can have a wildly different experience of life in an ADHD-kinda way; that is totally a thing.

A completely equivalent objection is that I don’t think “neurotypicality” is especially “real” — I am reluctant to feed the notion that there is a platonic ideal out there from which we might deviate, instead of a more plausible world of broken, maladapted apes trying to run a civilisation. High modernity’s preference to classify things into normal and aberrant looks threadbare, when it comes to ADHD. You do you though, if you feel otherwise.

If the line between treatment and enhancement is this thin, we are already a society practising widespread cognitive augmentation. The diagnostic threshold moves over time, prevalence estimates vary by a factor of three across countries, and the drugs work on a continuum. In a political economy dominated by the war for salience, ADHD might be a risk factor for exploitation by the attention economy, an adaptive response to it, or — per the transnormalists — a convenient diagnostic gateway through which a society that needs more sustained attention can purchase it one prescription at a time. Whatever, man, these all seem reasonable to me. If amphetamines make life better, this seems fine to me.

Bring on nootropics. I for one welcome our promethean reinvention of the human condition.

10 Implications for learning

We tend to think about attention, in part, as a learning mechanism now. What does ADHD say about the human learning mechanism?

Pollak (2023):

Indeed, children with ADHD needed more trials to reach the criterion, suggesting poor learning. However, they needed more trials under both the constant and the partial reinforcement conditions. On the extinction phase, children with ADHD made fewer previously-rewarded choices, suggesting faster extinction, as predicted by the DTD.

The combination of slow learning and fast extinction might seem surprising, as participants must learn the new rewards regimen in the extinction phase. Conversely, this combination is understandable; forgetting something one did not learn well is easier. Therefore, the results suggest that children with ADHD have a fundamental learning deficit, which is independent of whether the reinforcement is constant or partial.

See also (Addicott et al. 2021; Hulsbosch et al. n.d.; Nissan et al. 2023; Smith et al. 2022).

11 Incoming

• The ADHD Body Double: A Unique Tool for Getting Things Done - ADDA - Attention Deficit Disorder Association / Body doubling on Wikipedia

• An interesting coaching project: Off Piste ADHD Coaching – working with curious humans to discover what happens next

• John Cassidy, The Attention-Deficit-Disorder Economy

• Laura Basu, Attention deficit disorder, the anticapitalist condition

• Sluggish - by Jesse Meadows

  […]this is a newsletter where I write about disability and capitalism, covering the politics of mental health, the history of popular psychology, and the philosophies of living outside the norm.

• Neurodivergent Spoon Theory

• ADHD: Bottom-Up Triggers of Social Anxiety | Psychology Today

12 References