Stimulants and the Long Run: What ADHD Medication Does to Life Outcomes

The question behind the contradiction

Two literatures about the same drug point in opposite directions, and the contradiction is the story. In randomized trials, stimulant medication beats placebo reliably and by a wide margin within weeks: for methylphenidate the standardized effect on clinician-rated ADHD symptoms is about -0.8, and every first-line drug beats placebo on some measure. Yet the longest randomized trial on record, the famous MTA study, found that the advantage had evaporated by the end of a fourteen-month treatment phase and had not returned at three-year or eight-year follow-up. From that literature the natural conclusion is pessimistic: stimulants treat symptoms now and change nothing that lasts. Then one reads the registry studies, where medicated periods carry 15% to 40% lower rates of criminality, transport accidents, self-harm, and even death, and the conclusion flips.

The contradiction dissolves once the question is stated precisely. "Does medication help long-term?" bundles together two different kinds of outcome. There are developmental outcomes — symptom severity, executive function, academic attainment — where the payoff would have to come from the drug changing how a brain matures or how a person learns. And there are event outcomes — a criminal conviction, a car crash, a suicide attempt, an accidental death — where the relevant mechanism is momentary attention and impulse control at a specific moment. Stimulants are very good at the second and, on the evidence, do not deliver the first. The rest of this brief separates the two, weighs how much of the registry benefit is real, and states what would change the picture.

The randomized anchor, and why it is weaker than it looks

The MTA trial is the empirical center of gravity because it is the only large randomized study of ADHD treatment with a long follow-up. It randomized 579 children aged seven to just under ten with combined-type ADHD to fourteen months of medication management, intensive behavioral therapy, both, or routine community care [1]. During treatment, medication and combined care beat behavioral therapy and community care on symptom ratings — the standard acute result. But when the randomly assigned groups were re-examined two years after the controlled phase, they no longer differed on any primary measure [2], and at eight years the groups still did not differ on symptoms, grades, arrests, or psychiatric hospitalizations [3]. What predicted later functioning was the early symptom trajectory, not the treatment arm. By young adulthood, extended medication use in the naturalistic follow-up was associated with suppression of adult height but not with any reduction in symptom severity [4].

Two lessons follow, and both are routinely overstated. The first is that the MTA is not an argument that stimulants "do not work." It is an argument that the durable advantage of starting medication in childhood, as measured by symptoms and standard functioning, is absent once treatment intensity converges across groups — which it did, because community-treated children were also medicated and behavioral-therapy children later started taking pills. The trial randomized the initial strategy, not lifelong adherence. The second lesson is subtler: randomized evidence establishes what the average protocol does under ideal adherence, and it is silent about the registry outcomes that matter most, because a fourteen-month trial cannot count car crashes and convictions over a lifetime.

Short-run efficacy is not in dispute

It is worth being clear about the baseline, because a brief that opens with null long-run findings can read as though the drugs barely work. They work, acutely, on symptoms. A network meta-analysis of 133 double-blind randomized trials found that amphetamines, methylphenidate, and atomoxetine all beat placebo on clinician- and, where measured, teacher-rated ADHD symptoms in children and adolescents, with methylphenidate producing a standardized mean difference of -0.78 by clinician rating [5]. Long-term, a meta-analysis of chronic dosing found that methylphenidate and atomoxetine both improved executive-function measures — reaction time, attention, inhibition, working memory — with small-to-medium effects, the first pooled demonstration that the cognitive effects persist beyond a single dose [6]. The dispute is therefore not efficacy but translation: whether those acute and cognitive gains accumulate into a better life.

Crime and antisocial behavior

The first place the registry literature pushed back against the MTA's pessimism was criminality. Using Swedish national registers, a 2012 study compared the rate of criminal conviction during medicated periods with the rate for the same people during unmedicated periods, across 25,656 individuals. The conviction rate was 32% lower during medication for men (hazard ratio 0.68) and 41% lower for women (0.59), and the direction survived sensitivity analyses across drug types and offense categories [7]. The within-person comparison — the same person, medicated versus not — was the key design choice, because it removes stable differences between people who take medication and those who do not.

That design is not a panacea, and the same literature shows why. A Swedish study of suicidal behavior found the association reversed depending on the comparison: at the population level, medication appeared to raise the rate of suicide-related events by 31% (hazard ratio 1.31), but within the same individuals the rate was lower during treatment (0.89), and among stimulant users lower still (0.81) [8]. The population-level result reflected who gets medicated — sicker patients, at the moment of greatest crisis — rather than a harmful drug. The within-person estimate is the more credible one, which supports the benefit story, but the magnitude and even the sign of these estimates are visibly fragile to the choice of comparison group. That fragility is the central caution of this entire literature.

Accidents, injuries, and death

The event outcomes with the clearest mechanism — momentary inattention producing a crash or a fall — show the most consistent benefit. In 17,408 Swedish adults with ADHD, medication was associated with a 58% lower rate of serious transport accidents in men (hazard ratio 0.42), though the estimate in women was not statistically significant [9]. In children and adolescents, a within-person study found medication associated with an 15% lower rate of any unintentional injury (0.85) and a striking 73% lower rate of traumatic brain injury (0.27) [10]. A Quebec population-based cohort of 217,192 people aged 24 and under reported lower all-cause mortality (adjusted hazard ratio 0.61) and fewer emergency or hospital admissions for injury during medicated periods [11].

Mortality is the endpoint where the stakes are highest and the confounding is most severe, so the strongest design matters most. A Swedish cohort of 148,578 individuals used target-trial emulation — explicitly modeling the hypothetical experiment of starting versus not starting medication at diagnosis — and found lower two-year all-cause mortality in the initiators (hazard ratio 0.79), driven by a reduction in unnatural deaths such as accidents, suicide, and poisoning (0.75), with no significant difference in natural-cause deaths [12]. The absolute numbers keep the finding honest: 39.1 deaths per 10,000 in the initiation group versus 48.1 per 10,000 in the non-initiation group, a risk difference of about 9 per 10,000 over two years.

The largest single test came in 2025. A target-trial emulation of five outcomes over two years, in 148,581 people, found lower rates of first suicidal behaviors (rate ratio 0.83), substance misuse (0.85), transport accidents (0.88), and criminality (0.87) among those who started treatment — but no significant reduction in accidental injuries (0.98) [13]. A companion self-controlled analysis across three eras of rising prescription did not find that the associations weakened as medication use became more common, which argues against the benefits being an artifact of who seeks treatment early [14].

Substance use: the feared harm that did not appear

One long-standing worry — that treating childhood ADHD with stimulants primes the brain for later substance abuse — has been comparatively well tested, and the answer is reassuring. In the MTA cohort followed to a mean age of 25, neither current nor cumulative stimulant treatment predicted later heavy drinking, marijuana use, cigarette smoking, or other substance use, after adjustment for the developmental trends in substance use and for dynamic confounders [15]. The Swedish registry signal likewise points to lower substance-misuse rates during treatment [13, 16]. The most defensible reading is that stimulants neither cause nor prevent later substance-use disorder; the feared harm is not there, and the protection is modest.

Schooling: the near-null result

If any long-run benefit should show up, it is in school. It mostly does not. A Norwegian target-trial emulation of 8,548 children diagnosed with ADHD estimated the intention-to-treat effect of starting medication on national test scores and found effects that were statistically detectable but tiny — standardized mean differences of 0.037 for English, 0.063 for numeracy, and 0.071 for reading — which the authors judged not clinically relevant [17]. A second Norwegian study of 6,255 children went further, using naturally occurring variation in prescribing practice as an instrument: the ordinary-least-squares estimate suggested a small positive grade-point-average effect, but the instrumental-variable estimate, which better addresses residual confounding, found no causal effect on grades, test scores, or exemption rates [18]. Two independent designs, two nulls on the outcome families that matter most to families. Stimulants improve classroom behavior in the short term; they do not, on this evidence, make anyone learn more.

The safety ledger: growth, blood pressure, and the heart

Long-term use has costs, and they are better quantified than the benefits. The clearest is cardiovascular. A Swedish case-control study of 278,027 people with ADHD found that longer cumulative medication use was associated with higher odds of cardiovascular disease — about 4% per year of use (adjusted odds ratio 1.04), rising to 27% higher odds at three to five years of use, with hypertension showing the largest signal (nearly 80% higher odds beyond five years) [19]. The effect was present in children and youth as well as adults. This is a genuine safety finding and the strongest argument for cardiovascular monitoring on treatment.

Growth is a smaller story than the MTA suggested, at least over two years. The ADDUCE study, a prospective controlled study of 1,410 children across five European countries, found no evidence that two years of methylphenidate reduced height velocity (the standardized difference was -0.07 and not significant) or increased psychiatric or neurological adverse events, though pulse rate and blood pressure were modestly higher and no serious adverse events occurred [20]. The contrast with the MTA's height finding is instructive: the MTA compared naturalistic subgroups with different adherence, while ADDUCE compared groups by intention, with propensity adjustment. The MTA's height suppression may be partly real and partly the residue of who keeps taking medication for years.

Why the registry benefits are probably real but not as large as they look

Every observational estimate in this brief is vulnerable to four failure modes, and naming them keeps the conclusion calibrated.

• Confounding by indication. People are medicated because they are symptomatic, so "medicated periods" select for acuity. This is the force that made suicidal behavior look harmful between people (1.31) and protective within them (0.89) [8].
• The healthy-adherer effect. People who take medication consistently also tend to sleep, attend appointments, and avoid risk more. Adherence is a proxy for self-management.
• Immortal-time and outcome-dependent exposure. If follow-up time during which a person cannot yet have the outcome is counted as untreated, the untreated arm looks worse.
• Time-varying confounding. Within-person designs fix stable factors but not the dynamic ones — a crisis that both stops medication and precipitates an event.

Target-trial emulation [12, 13] addresses these more convincingly than naive cohort analysis, which is why its mortality and crime results carry more weight than the earlier within-person estimates. But emulation is still not randomization: it removes modeled confounding, and the residual is unknown. The Quebec mortality estimate of 0.61 [11], far larger than the Swedish target-trial estimate of 0.79 [12], is a useful detector: designs that control confounding less well report bigger benefits. The truth is likely at or below the more conservative end.

Two different kinds of outcome

The pattern across all of this evidence is not random. Medication changes event outcomes — crime, crashes, injuries, deaths — where a modest improvement in attention at the wrong moment prevents a discrete catastrophe. It does not change developmental outcomes — symptoms, learning, attainment — where the underlying ADHD trajectory and the ordinary mechanics of schooling dominate. A drug that sharpens attention for a few hours a day can plausibly prevent a crash or an impulsive arrest without shifting a grade-point average or extinguishing a symptom that is defined as an enduring pattern. That is a coherent mechanistic story, and it is what the data show. It also means the framing "does medication work long-term?" has no single answer: for catastrophes, probably yes; for symptoms and school, evidently no.

Calibrated verdict

• Very high confidence (~95%) that stimulants reduce ADHD symptoms versus placebo in the short term and improving executive-function measures with chronic dosing [5, 6].
• High confidence (~85%) that long-term medication does not normalize ADHD symptoms or produce durable symptom remission, and does not improve academic attainment, across randomized and target-trial evidence [2, 3, 4, 17, 18].
• Moderate-to-high confidence (~70-75%) that medicated periods genuinely reduce rates of criminality, transport accidents, self-harm, and unnatural death, with a plausible mechanistic basis, while accepting that residual confounding inflates the observed magnitudes [7, 9, 10, 12, 13].
• Moderate-to-high confidence (~75%) that long-term use carries a small but real cardiovascular cost, largest for hypertension, that justifies routine monitoring [19, 20].
• High confidence (~85%) that stimulant treatment does not increase the risk of later substance-use disorder [15, 16].

Reversal conditions

This reading would change if any of the following appeared. First, a long-horizon randomized or preference-based trial showing a durable symptom or attainment benefit that survives when treatment intensity is held equal — the MTA's central ambiguity. Second, a Mendelian-randomization or sibling-comparison study isolating the medication effect from the treatment-seeking that drives it, and finding the registry benefits persist at full size. Third, replication of the MTA height finding in intention-to-treat designs, which would move the safety ledger against medication. Fourth, evidence that the registry benefits weaken as prescribing widens — the opposite of what the era analysis found [14] — which would suggest the early benefits were selection rather than pharmacology.

The study that would settle it

The decisive design does not yet exist, and it is well defined: a large randomized or strong quasi-experimental comparison of sustained medication versus high-quality non-pharmacological care, with registry-linked outcomes over a decade or more — criminality, transport accidents, cardiovascular events, attainment — and with equalized non-drug support, so that the contrast isolates the drug. No such trial exists, in part because randomizing people away from available treatment is ethically fraught and because the outcomes accrue over years. The closest feasible substitute is a within-sibship or instrumental-variable design powered for the event outcomes, conducted in a system with complete register linkage and validated prescribing variation. Until one of these lands, the best-supported summary is the uncomfortable one: stimulants reliably relieve ADHD symptoms and reliably do not eliminate the disorder, while probably preventing a meaningful share of the discrete catastrophes that punctuate it — at the price of a small cardiovascular risk that monitoring can contain.

References

1. The MTA Cooperative Group. A 14-month randomized clinical trial of treatment strategies for attention-deficit/hyperactivity disorder. Archives of General Psychiatry. 1999;56(12):1073-1086. doi:10.1001/archpsyc.56.12.1073
2. Jensen PS, Arnold LE, Swanson JM, et al. 3-year follow-up of the NIMH MTA study. Journal of the American Academy of Child & Adolescent Psychiatry. 2007;46(8):989-1002. doi:10.1097/CHI.0b013e3180686d48
3. Molina BSG, Hinshaw SP, Swanson JM, et al. The MTA at 8 years: prospective follow-up of children treated for combined-type ADHD in a multisite study. Journal of the American Academy of Child & Adolescent Psychiatry. 2009;48(5):484-500. doi:10.1097/CHI.0b013e31819c23d0
4. Swanson JM, Arnold LE, Molina BSG, et al. Young adult outcomes in the follow-up of the multimodal treatment study of attention-deficit/hyperactivity disorder: symptom persistence, source discrepancy, and height suppression. Journal of Child Psychology and Psychiatry. 2017;58(6):663-678. doi:10.1111/jcpp.12684
5. Cortese S, Adamo N, Del Giovane C, et al. Comparative efficacy and tolerability of medications for attention-deficit hyperactivity disorder in children, adolescents, and adults: a systematic review and network meta-analysis. The Lancet Psychiatry. 2018;5(9):727-738. doi:10.1016/S2215-0366(18)30269-4
6. Isfandnia F, El Masri S, Radua J, et al. The effects of chronic administration of stimulant and non-stimulant medications on executive functions in ADHD: a systematic review and meta-analysis. Neuroscience & Biobehavioral Reviews. 2024;162:105703. doi:10.1016/j.neubiorev.2024.105703
7. Lichtenstein P, Halldner L, Zetterqvist J, et al. Medication for attention deficit-hyperactivity disorder and criminality. New England Journal of Medicine. 2012;367(21):2006-2014. doi:10.1056/NEJMoa1203241
8. Chen Q, Sjolander A, Runeson B, et al. Drug treatment for attention-deficit/hyperactivity disorder and suicidal behaviour: register based study. BMJ. 2014;348:g3769. doi:10.1136/bmj.g3769
9. Chang Z, Lichtenstein P, D'Onofrio BM, et al. Serious transport accidents in adults with attention-deficit/hyperactivity disorder and the effect of medication. JAMA Psychiatry. 2014;71(3):319-325. doi:10.1001/jamapsychiatry.2013.4174
10. Ghirardi L, Chen Q, Chang Z, et al. Use of medication for attention-deficit/hyperactivity disorder and risk of unintentional injuries in children and adolescents with co-occurring neurodevelopmental disorders. Journal of Child Psychology and Psychiatry. 2020;61(2):140-147. doi:10.1111/jcpp.13136
11. Vasiliadis HM, Lunghi C, Rahme E, et al. ADHD medications use and risk of mortality and unintentional injuries: a population-based cohort study. Translational Psychiatry. 2024;14(1):128. doi:10.1038/s41398-024-02825-y
12. Li L, Zhu N, Zhang L, et al. ADHD pharmacotherapy and mortality in individuals with ADHD. JAMA. 2024;331(10):850-860. doi:10.1001/jama.2024.0851
13. Zhang L, Zhu N, Sjolander A, et al. ADHD drug treatment and risk of suicidal behaviours, substance misuse, accidental injuries, transport accidents, and criminality: emulation of target trials. BMJ. 2025;390:e083658. doi:10.1136/bmj-2024-083658
14. Li L, Coghill D, Sjolander A, et al. Increased prescribing of attention-deficit/hyperactivity disorder medication and real-world outcomes over time. JAMA Psychiatry. 2025;82(8):830-837. doi:10.1001/jamapsychiatry.2025.1281
15. Molina BSG, Kennedy TM, Howard AL, et al. Association between stimulant treatment and substance use through adolescence into early adulthood. JAMA Psychiatry. 2023;80(9):933-941. doi:10.1001/jamapsychiatry.2023.2157
16. Chang Z, Lichtenstein P, Halldner L, et al. Stimulant ADHD medication and risk for substance abuse. Journal of Child Psychology and Psychiatry. 2014;55(8):878-885. doi:10.1111/jcpp.12164
17. Varnet Perez T, Overgaard KR, Frigessi A, et al. Long-term effect of pharmacological treatment on academic achievement of Norwegian children diagnosed with ADHD: a target trial emulation. International Journal of Epidemiology. 2025;54(2):dyaf010. doi:10.1093/ije/dyaf010
18. Lyhmann I, Widding-Havneraas T, Markussen S, et al. Effect of pharmacological treatment of attention-deficit hyperactivity disorder on academic performance: registry-based population study. The British Journal of Psychiatry. 2026:1-8. doi:10.1192/bjp.2026.10754
19. Zhang L, Li L, Andell P, et al. Attention-deficit/hyperactivity disorder medications and long-term risk of cardiovascular diseases. JAMA Psychiatry. 2024;81(2):178-187. doi:10.1001/jamapsychiatry.2023.4294
20. Man KKC, Hage A, Banaschewski T, et al. Long-term safety of methylphenidate in children and adolescents with ADHD: 2-year outcomes of the Attention Deficit Hyperactivity Disorder Drugs Use Chronic Effects (ADDUCE) study. The Lancet Psychiatry. 2023;10(5):323-333. doi:10.1016/S2215-0366(23)00042-1