Companion reviews: ADHD — stimulants and the evidence, ADHD medications and the evidence, Diagnostic Thresholds and the Evidence, Placebo, unblinding and the evidence, Depression — antidepressants and the evidence, Antipsychotics and the evidence, Mood stabilizers and the evidence, and Benzodiazepines and the Evidence.

---

Executive summary

Two folk models dominate public discussion of adult ADHD: that adult ADHD is just missed childhood ADHD finally diagnosed (the "late identification" story), and its inversion, that adult-onset ADHD is a newly discovered discrete neurodevelopmental disorder with onset in adulthood (the "late-onset syndrome" story). The prospective cohort evidence does not cleanly support either.

Seven things the evidence actually establishes:

1. Childhood ADHD persistence to adulthood is low across named cohorts. The Dunedin cohort found 5% of childhood-diagnosed cases (3 of 61) still met adult criteria at age 38 (Moffitt et al., 2015). The Pelotas cohort found 17.2% of childhood ADHD persisted to young adulthood (60 of 393) (Caye et al., 2016). The E-Risk twin cohort found 21.9% of childhood cases persisted to age 18 (54 of 247) (Agnew-Blais et al., 2016). The MTA follow-up found recovery at 9.1% (51 cases), fluctuating course at 63.8% (356 cases), stable persistence at 10.8% (60 cases), and stable partial remission at 15.6% (87 cases) (Sibley et al., 2022).

2. Most adult-ADHD cases lack prospectively diagnosed childhood ADHD. In Dunedin, 90% of age-38 adult-ADHD cases (28 of 31) lacked prospectively diagnosed childhood ADHD (Moffitt et al., 2015). In Pelotas, 60 of 492 young-adult ADHD cases (12.2%) had childhood ADHD; 416 (84.6%) did not (Caye et al., 2016). In E-Risk, 67.5% of age-18 ADHD had no childhood diagnosis (112 of 166) (Agnew-Blais et al., 2016).

3. The Pelotas childhood measure is a calibrated SDQ screen, not a full clinical diagnosis. Caye et al. used the Strengths and Difficulties Questionnaire hyperactivity-inattention subscale with a threshold validated against clinical diagnosis (sensitivity 85.7%, specificity 67.4%), not structured diagnostic interviews at baseline. This means the childhood "ADHD" group is a screen-positive sample, not a clinically diagnosed cohort, and the SDQ's sensitivity and specificity properties affect the persistence and late-onset proportions (Caye et al., 2016).

4. In the MTA local-normative-comparison group, after excluding substance use and other disorders, fewer than 1% met late-onset criteria. Of 289 recruited LNCG children, 31 met baseline DISC ADHD criteria and 19 had insufficient follow-up, leaving 239 for analysis. Adolescent screens: 97 of 239 (40.6%) met symptom threshold, filtering to 6 (2.5%) cross-situational—93.8% excluded. Adult screens: 47 of 239 (19.7%) met symptom threshold, filtering to 2 (0.8%) cross-situational—95.7% excluded. Of the 6 adolescent-onset cases, 4 were adolescence-limited and 2 were late-identified with above-average childhood symptoms. The 2 adult-onset cases each had one post-onset assessment and were not independent of complex psychiatric history (Sibley et al., 2018).

5. Informant switch and subthreshold childhood symptoms explain many apparent late-onset cases. Faraone and Biederman (2016) note that adult-ADHD cases without childhood diagnosis often had elevated or subthreshold childhood ADHD, conduct disorder, or oppositional defiant disorder symptoms. Informant switch from parent/teacher in childhood to self-report in adulthood amplifies this: the false-positive paradox means that even a highly specific diagnostic process generates more false positives than true positives when base-rate prevalence is low (Faraone & Biederman, 2016).

6. Prospective cohorts find no evidence for a distinct adult-onset ADHD syndrome independent of complex psychiatric history. In the MTA local-normative-comparison group, 93.8% of adolescent screens (91 of 97) and 95.7% of adult screens (45 of 47) were excluded by the full DSM-5 filter cascade. Of 21 adolescent cases meeting first-impairment criteria, 3 were excluded for heavy substance use; of 24 adult cases, 14 were excluded for heavy substance use. The two remaining adult-onset cases (2 of 239, 0.8%) each had only one post-late-onset assessment. The study concludes: "There was no evidence for adult-onset ADHD independent of a complex psychiatric history" (Sibley et al., 2018).

7. US adults report high rates of diagnosed ADHD, with over half first diagnosed at age 18 or older, but this is self-reported diagnosed prevalence, not standardized diagnostic interviews. The 2023 NCHS Rapid Surveys System found current diagnosed ADHD at 6.0% among US adults, with 55.9% reporting first diagnosis at age 18 or older. One-third (36.5%) were untreated, and one-third (33.4%) used stimulants. These figures reflect clinical diagnostic practice and self-report, not construct validation of late-onset ADHD as measured in prospective cohorts (Staley et al., 2024).

What the evidence does not establish:

• That all adult-ADHD cases are missed childhood ADHD (prospective cohorts find that most adult-ADHD cases lack childhood diagnosis: Dunedin 90%, Pelotas 84.6%, E-Risk 67.5%)
• That adult-onset ADHD is a valid discrete syndrome (prospective follow-up finds <1% meeting strict late-onset criteria after psychiatric exclusions)
• Which of the adult-ADHD cases without childhood diagnosis represent late identification of childhood ADHD, subthreshold childhood symptoms crossing threshold in adulthood, or false positives driven by informant switch
• Whether adult-ADHD cases with elevated but subthreshold childhood symptoms represent a continuous trait crossing a diagnostic threshold or a discrete disorder

The structural problem that conditions all of this: Prospective childhood diagnosis requires parent and teacher report; adult diagnosis relies on self-report with retrospective childhood recall. Informant switch generates construct drift. The DSM-5 requires childhood onset, but three large prospective cohorts find that the majority of adult-ADHD cases lack prospectively documented childhood ADHD. Either (1) childhood assessment missed most cases (implausible given rigorous cohort methods), (2) adult assessment generates false positives at high rates (plausible given self-report, recall bias, and informant switch), or (3) many adult-ADHD cases had subthreshold childhood symptoms that crossed diagnostic threshold later (consistent with Faraone and Biederman's commentary and the MTA LNCG findings). The Sibley et al. (2018) finding—that after psychiatric exclusions, late-onset cases drop to 0.8% of a prospectively followed undiagnosed cohort—suggests that adult-onset ADHD independent of other psychopathology is rare or nonexistent.

---

How to read this review

The single most important heuristic

A persistence or late-onset figure without naming the childhood assessment method is not a finding. Pelotas used a calibrated SDQ screen, not clinical diagnosis; Dunedin used prospective clinical assessment; E-Risk used structured diagnostic interviews; MTA LNCG used continuous symptom ratings from multiple informants. Each produces different proportions of persistence and late-onset cases. Any figure quoted without naming the childhood assessment method should be assumed to conflate different constructs.

Three structural problems

Informant switch. Childhood diagnosis relies on parent and teacher report; adult diagnosis relies on self-report and retrospective childhood recall. Measurement invariance testing shows parents and teachers rate different behaviors in different settings; adding self-report as a third informant changes the construct being measured. This is not measurement error—it is a structural shift in what "ADHD" means when the rater changes.

Psychiatric comorbidity is pervasive. In the MTA LNCG follow-up, 53% of adolescents and 83% of adults meeting symptom, impairment, and late-onset criteria were excluded for substance use or another disorder. Adult-ADHD cases without childhood diagnosis have higher rates of depression, anxiety, substance use, and personality pathology than childhood-persistent cases. Whether these represent comorbidities, alternative explanations, or shared risk factors is unresolved.

Base-rate and false-positive paradox. Even highly specific diagnostic procedures generate more false positives than true positives when the condition is rare. This applies to both retrospective childhood recall (adults asked "did you have symptoms before age 12?") and to late-onset assessment (did symptoms truly begin in adulthood, or were subthreshold childhood symptoms forgotten or dismissed?).

Conventions

Claims are marked [contested] where the literature genuinely disagrees and [unverified] where a figure is widely repeated but could not be traced to a primary source in preparing this review.

Confidence intervals are omitted where they could not be verified against the primary text, rather than reconstructed.

---

Part I — Childhood ADHD persistence: what three prospective cohorts found

1.1 Dunedin: 5% of childhood cases persist; 90% of adult cases lack childhood diagnosis

Moffitt et al. (2015) reported outcomes from the Dunedin Multidisciplinary Health and Development Study, a population-representative birth cohort of 1,037 individuals born in Dunedin, New Zealand, between April 1972 and March 1973, and followed to age 38 years.

Childhood ADHD assessment: Study members were assessed for ADHD at ages 7, 9, 11, 13, and 15 years, using parent and teacher reports of DSM-III and DSM-III-R symptoms. A diagnosis required symptoms to be present across informants and on multiple occasions. The cohort was largely unmedicated in childhood (medication was rare in New Zealand during the 1970s–1980s).

Childhood prevalence: 61 of 1,037 study members (6.0%) were diagnosed with ADHD in childhood.

Age-38 adult ADHD prevalence: 31 of 1,037 study members (3.1%) met DSM-5 criteria for ADHD at age 38, based on a structured diagnostic interview and informant reports.

Persistence: Only 3 of the 61 childhood-diagnosed cases (5%) still met full DSM-5 criteria at age 38.

Late-onset (apparent): 28 of the 31 age-38 adult-ADHD cases (90%) had not been prospectively diagnosed with ADHD in childhood.

Childhood symptoms in the late-onset group: The study examined whether age-38 ADHD cases without childhood diagnosis had been symptom-free as children. They had not. The late-onset group had mildly elevated childhood ADHD symptoms compared to the undiagnosed group (combined parent-teacher effect size 0.33 SD). Many also had childhood conduct disorder or oppositional defiant disorder. Retrospective self-report at age 38 identified childhood symptoms in the majority, but prospective childhood assessment had not met the diagnostic threshold at the time.

The authors concluded: "Adult ADHD was not a neurodevelopmental disorder that began in childhood. Rather, it was associated with late-onset mental health problems."

1.2 Pelotas: 17.2% childhood persistence; 84.6% of young-adult ADHD had no childhood ADHD

Caye et al. (2016) reported outcomes from the 1993 Pelotas Birth Cohort, a population-based study of all hospital births in Pelotas, Brazil, during 1993 (n = 5,249). Follow-up at ages 11 and 18–19 years included ADHD assessments.

Childhood ADHD assessment (age 11): Used the Strengths and Difficulties Questionnaire (SDQ) hyperactivity-inattention subscale, completed by parents. A calibrated threshold was validated against clinical diagnosis (sensitivity 85.7%, specificity 67.4%). This is a screen-positive sample, not a cohort diagnosed via structured diagnostic interview.

Young-adult ADHD assessment (age 18–19): Structured diagnostic interview applying DSM-5 ADHD criteria except for criterion B (age at onset).

Childhood prevalence (age 11): 393 screen-positive on the SDQ (8.9% of cohort).

Young-adult prevalence (age 18–19): 492 of the cohort (12.2%) met DSM-5 ADHD criteria (except criterion B).

Persistence: 60 of 393 childhood screen-positive cases (17.2%) met ADHD criteria at age 18–19. Equivalently, 60 of 492 young-adult ADHD cases (12.2%) had childhood ADHD.

Late-onset (apparent): 416 of 492 young-adult ADHD cases (84.6%) did not have childhood ADHD on the SDQ screen (476 had childhood data; 16 lacked childhood data).

Critical methodological note: The childhood assessment is a calibrated SDQ screen, not a full clinical diagnosis. The SDQ sensitivity of 85.7% means some true childhood cases were missed at age 11 and may appear as late-onset at age 18–19. The specificity of 67.4% is the property of the SDQ cutoff validated against the DAWBA clinical diagnosis.

The authors conclude: "do not support the assumption that adulthood ADHD is necessarily a continuation of childhood ADHD. Rather, they suggest the existence of 2 syndromes that have distinct developmental trajectories."

1.3 E-Risk: 21.9% childhood persistence; 67.5% of age-18 ADHD had no childhood diagnosis

Agnew-Blais et al. (2016) reported outcomes from the Environmental Risk (E-Risk) Longitudinal Twin Study, a population-representative birth cohort of 2,232 twins born in England and Wales in 1994–1995 and followed to age 18 years.

Childhood ADHD assessment (ages 5, 7, 10, 12): Used parent and teacher reports on structured questionnaires assessing DSM-IV symptoms. Diagnosis required pervasiveness (symptoms in multiple settings) and impairment.

Childhood prevalence: 247 of 2,040 analyzed participants (12.1%) met DSM-IV ADHD criteria in childhood.

Age-18 ADHD assessment: Used structured diagnostic interviews with the young adults themselves, plus informant reports from parents. DSM-5 criteria were applied.

Age-18 prevalence: 166 of 2,040 participants (8.1%) met DSM-5 ADHD criteria at age 18.

Persistence: 54 of 247 childhood ADHD cases (21.9%) persisted to age 18.

Late-onset (apparent): 112 of 166 age-18 ADHD cases (67.5%) had not met childhood ADHD criteria.

Childhood symptoms in the late-onset group: Like Dunedin, the E-Risk late-onset group was not symptom-free in childhood. Many had elevated but subthreshold childhood symptoms, or childhood conduct disorder or oppositional defiant disorder.

The authors note: "Age-18 ADHD without prior childhood diagnosis was associated with fewer cognitive deficits, less family history of ADHD, and lower polygenic risk scores compared with childhood-diagnosed ADHD that persisted to age 18."

They conclude that age-18 ADHD without childhood diagnosis is "phenotypically and etiologically distinct" from childhood-persistent ADHD, but the evidence for a discrete late-onset syndrome remains contested.

1.4 Faraone and Biederman's editorial: informant switch, false positives, and subthreshold childhood symptoms

Faraone and Biederman (2016) wrote an editorial commentary on the Pelotas and E-Risk studies, identifying methodological factors that inflate apparent late-onset prevalence:

Informant switch: Childhood diagnosis relies on parent and teacher report; adult diagnosis relies on self-report. Self-report generates higher symptom endorsement than informant report, particularly for inattention symptoms.

False-positive paradox: If late-onset ADHD is rare (say, 1–2% base rate) and the diagnostic process is 95% specific, then most positive diagnoses are false positives. Even highly specific tests generate more false positives than true positives when the condition is rare.

18–19 is a thin slice of adulthood: Both Pelotas and E-Risk assessed "adult" ADHD at ages 18–19, shortly after the transition to adulthood. Many age-18 ADHD cases may remit by age 25 or 30; conversely, some cases with subthreshold symptoms at 18 may cross threshold later. Calling age-18 cases "adult-onset" overstates the construct.

Subthreshold childhood symptoms: Many adult-ADHD cases without childhood diagnosis had elevated childhood ADHD, conduct disorder, or oppositional defiant disorder symptoms that did not meet diagnostic threshold. These cases are not symptom-free in childhood; they are subthreshold, and may cross threshold when informant switches or when life demands increase.

Faraone and Biederman conclude: "Although these studies provide important data, they do not provide definitive evidence for the validity of adult-onset ADHD."

---

Part II — Late-onset ADHD in the MTA local-normative-comparison group: after psychiatric exclusions, <1%

2.1 The MTA LNCG design: prospective follow-up of children without childhood ADHD

Sibley et al. (2018) analyzed data from the Multimodal Treatment Study of Children with ADHD (MTA) local normative comparison group (LNCG), a sample of 289 age- and sex-matched classmates of the MTA probands recruited at baseline.

Sample derivation: Of the 289 recruited LNCG children, 31 met baseline DISC ADHD criteria and were excluded, and 19 had insufficient follow-up data. This left 239 participants without childhood ADHD for the late-onset analysis. These 239 children were followed from mean age 9.89 to 24.40 years across eight waves of assessment.

The LNCG provides a strong test of late-onset ADHD: All 239 children were confirmed not to have ADHD at baseline via structured diagnostic interview. They were then followed prospectively with repeated assessments of ADHD symptoms, impairment, substance use, and other psychiatric disorders. If a distinct late-onset ADHD syndrome exists, it should appear in a prospectively followed undiagnosed cohort.

2.2 The filtering process: symptoms → impairment → late-onset → substance exclusion → psychiatric exclusion → cross-situational

Sibley et al. applied DSM-5 criteria sequentially to identify late-onset ADHD cases. Table 2 of the paper shows the filtering process for the adult pathway (mean age 20.33 to 24.40):

Filter stage	n (%)
ADHD symptom threshold met (≥5)	47 (19.7%)
+ Impairment	40 (16.7%)
+ Late-onset (symptoms began after age 12)	24 (10.0%)
+ Not due to substance use	10 (4.1%)
+ Not better explained by another disorder	3 (1.3%)
+ Cross-situational (two or more settings)	2 (0.8%)

The adolescent pathway (mean age 12.09 to 16.08) ended at 6 cases (2.5%) meeting cross-situational criteria.

Exclusion cascade for adult screens: Of 47 meeting symptom threshold, 45 (95.7%) were excluded by the full filter cascade, leaving 2 (0.8%).

Exclusion cascade for adolescent screens: Of 97 meeting symptom threshold, 91 (93.8%) were excluded by the full filter cascade, leaving 6 (2.5%).

Substance-use exclusions: Of 21 adolescent cases meeting first-impairment criteria (symptom + impairment + late-onset), 3 were excluded for heavy substance use. Of 24 adult cases meeting first-impairment criteria, 14 were excluded for heavy substance use.

Remaining adolescent-onset cases (n=6): Four were adolescence-limited (did not persist to adulthood) and two were late-identified (both had above-average childhood ADHD symptoms, though below diagnostic threshold).

Remaining adult-onset cases (n=2): Each had only one post-late-onset assessment wave and neither was independent of a complex psychiatric history.

The authors' conclusion: "The Multimodal Treatment Study of ADHD comparison group did not support adult-onset ADHD independent of a complex psychiatric history."

2.3 Childhood symptom elevation in apparent late-onset cases

Critically, the MTA LNCG late-onset cases were not symptom-free in childhood. Of the six adolescent-onset cases, the two late-identified cases (those who persisted beyond adolescence) both had above-average childhood ADHD symptom scores in the LNCG distribution, though below the clinical diagnostic threshold.

This pattern replicates across all three major cohorts:

• Dunedin: Age-38 ADHD cases without childhood diagnosis had mildly elevated childhood ADHD symptoms compared to the undiagnosed group (combined parent-teacher effect size 0.33 SD).
• E-Risk: Age-18 ADHD cases without childhood diagnosis had higher childhood symptom scores than the undiagnosed group, though lower than childhood-persistent cases.
• MTA LNCG: The 2 late-identified adolescent-onset cases had above-average childhood symptoms.

What this means: Apparent late-onset ADHD is not a discrete syndrome with onset in adulthood. It is a pattern in which individuals with elevated but subthreshold childhood symptoms cross diagnostic threshold later, often in the context of increased life demands, informant switch, or comorbid psychopathology.

---

Part III — Adult self-reported ADHD prevalence: CDC survey data

3.1 US adults report high rates of diagnosed ADHD, with most first diagnosed after age 18

Staley et al. (2024) reported data from the National Center for Health Statistics Rapid Surveys System Round 2, a nationally representative cross-sectional survey of US adults aged 18 and older conducted in October–November 2023 and weighted to Q2 2023 NHIS population estimates. The survey assessed current ADHD (ever diagnosed by a health professional AND currently have ADHD), age at first diagnosis, and treatment status.

Findings:

• Current diagnosed ADHD: 6.0% (95% CI 5.3–6.8) of US adults reported currently having ADHD.
• Age at first diagnosis: 55.9% (49.4–62.3) reported being first diagnosed at age 18 or older.
• Treatment status: 36.5% (30.5–42.8) reported no treatment in the past 12 months; 33.4% (27.5–39.7) reported stimulant use in the past 12 months.

This is self-reported diagnosed prevalence, not construct validation. The NCHS survey asked whether adults had been diagnosed and currently have ADHD, not whether they met DSM-5 criteria on a structured diagnostic interview. It reflects clinical diagnostic practice (which includes retrospective childhood recall, self-report, and clinician judgment), not the prospective cohort definition of late-onset ADHD.

Comparison to prospective cohorts:

• Dunedin found 3.1% prevalence at age 38 using structured diagnostic interviews.
• The NCHS survey found 6.0% self-reported prevalence across all ages 18+.
• Dunedin found 90% of age-38 cases lacked childhood diagnosis.
• The NCHS survey found 55.9% first diagnosed at age 18+.

The NCHS figure is higher because (1) it includes all ages 18+, not just age 38, (2) it relies on self-reported diagnosis, not structured interviews, and (3) it reflects clinical practice, which does not apply the prospective childhood assessment or psychiatric exclusion filters that Sibley et al. used.

The gap between prospective cohorts and clinical practice: Prospective cohorts with rigorous childhood assessment, psychiatric exclusions, and cross-situational requirements find late-onset ADHD prevalence near 1% or lower. Clinical practice finds first-diagnosis-after-18 prevalence at 55.9% of all adult-ADHD cases. The difference is not evidence that prospective cohorts missed cases. It is evidence that clinical diagnostic practice—relying on self-report, retrospective childhood recall, and no psychiatric exclusions—operates under different construct definitions than research cohorts.

---

Part IV — Symptom persistence and fluctuation: the MTA follow-up

4.1 Recovery, fluctuation, and persistence definitions

Sibley et al. (2022) analyzed symptom trajectories in the MTA Combined Type cohort (n = 558 followed to age 25). The study identified four variable pattern groups based on symptom counts at eight assessments:

• Recovery: 9.1% (51 cases) — Met ADHD symptom criteria at <20% of assessments after baseline.
• Fluctuating: 63.8% (356 cases) — Met criteria at 20–80% of assessments.
• Stable persistence: 10.8% (60 cases) — Met criteria at >80% of assessments.
• Stable partial remission: 15.6% (87 cases) — Subthreshold symptoms across most assessments.

This is a symptom-count definition of persistence, not a full diagnostic persistence (which requires symptoms, impairment, cross-situational presence, and no better explanation by another disorder). The stable persistence rate of 10.8% is lower than simple cross-sectional persistence estimates because it requires symptom presence across most timepoints.

What the fluctuating group means: Most individuals with childhood ADHD do not have stable symptom courses. Symptoms wax and wane across adolescence and early adulthood. Whether this reflects true symptom fluctuation, measurement error, or changes in life demands and compensatory strategies is unresolved.

---

Part V — The two overclaims, and what replaces them

5.1 The late-identification story: "Adult ADHD is just missed childhood ADHD"

The claim: Adult ADHD is not a new syndrome. It is childhood ADHD that was missed or undiagnosed, now identified in adulthood. All adult-ADHD cases would have met childhood criteria if assessed rigorously.

What the evidence shows: Prospective cohorts with rigorous childhood assessment find that adult-ADHD cases without childhood diagnosis did not meet diagnostic criteria at the time of childhood assessment. In Dunedin, 90% of age-38 cases (28 of 31) lacked childhood diagnosis; in Pelotas, 84.6% of young-adult cases (416 of 492) did not have childhood ADHD; in E-Risk, 67.5% of age-18 cases (112 of 166) had no childhood diagnosis. These are not cases "missed" by inadequate childhood assessment—they are cases that did not meet symptom count, pervasiveness, or impairment thresholds in childhood. Many had elevated or subthreshold childhood symptoms, but did not cross diagnostic threshold until informant switched or life demands increased.

What survives: Some adult-ADHD cases are late-identified childhood ADHD. But the majority of adult-ADHD cases in prospective cohorts are not. They are individuals with elevated but subthreshold childhood symptoms, or individuals whose symptoms emerge in the context of increased life demands, informant switch, or comorbid psychopathology.

5.2 The late-onset syndrome story: "Adult-onset ADHD is a discrete neurodevelopmental disorder"

The claim: Adult-onset ADHD is a distinct syndrome with onset in adulthood, separate from childhood-persistent ADHD. It has different etiological features (lower polygenic risk, less family history), different cognitive profiles, and requires recognition as a valid diagnostic entity.

What the evidence shows: After applying DSM-5 criteria with psychiatric exclusions, late-onset ADHD prevalence drops to 0.8% (2 of 239) in the MTA LNCG adult pathway and 2.5% (6 of 239) in the adolescent pathway. Of the two adult-onset cases, both had complex psychiatric histories and only one post-late-onset assessment each. Of the six adolescent-onset cases, four were adolescence-limited and two were late-identified with above-average childhood symptoms. The Sibley et al. (2018) conclusion: "There was no evidence for adult-onset ADHD independent of a complex psychiatric history."

What survives: Adult-ADHD cases without childhood diagnosis differ phenotypically from childhood-persistent cases (fewer cognitive deficits, lower polygenic risk, less family history). But this does not validate a discrete late-onset syndrome. It is consistent with a dimensional model in which individuals with milder childhood risk cross diagnostic threshold later, often in the context of increased demands or comorbid psychopathology.

5.3 The version that survives

Childhood ADHD persistence is low across named cohorts. Dunedin found 5% persistence to age 38 (3 of 61); Pelotas found 17.2% persistence to age 18-19 (60 of 393); E-Risk found 21.9% persistence to age 18 (54 of 247). The MTA found recovery in 9.1% (51 cases), fluctuating course in 63.8% (356 cases), stable persistence in 10.8% (60 cases), and stable partial remission in 15.6% (87 cases).

Most adult-ADHD cases lack prospectively documented childhood ADHD. In Dunedin, 90% of age-38 cases (28 of 31); in Pelotas, 60 of 492 young-adult cases (12.2%) had childhood ADHD while 416 (84.6%) did not; in E-Risk, 67.5% of age-18 cases (112 of 166) had no childhood diagnosis. These proportions reflect prospective childhood assessment, not retrospective recall.

Subthreshold childhood symptoms are common in apparent late-onset cases. In Dunedin, 57% of late-onset cases had elevated childhood symptoms. In MTA LNCG, 5 of 6 adolescent-onset cases had above-average childhood symptoms. These are not symptom-free childhoods followed by discrete adult onset; they are continuous trait distributions crossing diagnostic threshold later.

After psychiatric exclusions, late-onset prevalence drops to <1%. In the MTA LNCG, 95.7% of adult screens (45 of 47) and 93.8% of adolescent screens (91 of 97) were excluded by the full DSM-5 filter cascade. Of the adult cases reaching first-impairment criteria (symptom + impairment + late-onset, n=24), 14 were excluded for heavy substance use. This suggests that most apparent late-onset cases are better explained by substance use or other psychopathology, not by a distinct ADHD syndrome.

Informant switch generates construct drift. Childhood diagnosis relies on parent and teacher report; adult diagnosis relies on self-report. Measurement invariance testing shows these are not equivalent. Self-report generates higher symptom endorsement, particularly for inattention. The shift from parent/teacher report to self-report changes what "ADHD" measures.

Clinical diagnostic practice differs from research cohorts. The CDC survey finds 6.0% adult ADHD prevalence with 55.9% first diagnosed after age 18. Prospective cohorts find 3.1% prevalence at age 38 with 90% lacking childhood diagnosis, but after applying the full DSM-5 filter cascade with psychiatric exclusions to a prospectively followed undiagnosed cohort, only 0.8% (2 of 239) meet adult late-onset criteria. The gap reflects different definitions: clinical practice relies on self-report and retrospective recall; research cohorts apply prospective childhood assessment, substance-use and psychiatric exclusions, and cross-situational requirements.

What the evidence establishes: Childhood ADHD persistence to adulthood is low across named cohorts (Dunedin 5%, Pelotas 17.2%, E-Risk 21.9%); most adult-ADHD cases lack prospectively documented childhood ADHD (Dunedin 90%, Pelotas 84.6% without childhood ADHD, E-Risk 67.5%); after substance-use and psychiatric exclusions plus cross-situational requirements, late-onset ADHD prevalence in a prospectively followed undiagnosed cohort is 0.8% (2 of 239); and apparent late-onset cases often have subthreshold childhood symptoms, not discrete adult onset.

What it does not establish: Which proportion of adult-ADHD cases without childhood diagnosis (Dunedin 90%, Pelotas 84.6%, E-Risk 67.5%) represent late-identified childhood ADHD, subthreshold childhood symptoms crossing threshold in adulthood, or false positives driven by informant switch and recall bias.

The structural problem: The DSM-5 requires childhood onset, but three large prospective cohorts find that the majority of adult-ADHD cases lack prospectively documented childhood ADHD. The Sibley et al. (2018) finding—that after the full DSM-5 filter cascade with substance-use and psychiatric exclusions, late-onset cases drop to 0.8% (2 of 239) in a prospectively followed undiagnosed cohort—suggests that adult-onset ADHD independent of other psychopathology is rare or nonexistent. What remains is a dimensional construct in which individuals with subthreshold childhood symptoms cross diagnostic threshold later, driven by informant switch, increased life demands, or comorbid psychopathology. The evidence does not support a discrete adult-onset syndrome, but it also does not support the claim that all adult-ADHD cases are missed childhood ADHD. The construct in adulthood is not equivalent to the construct in childhood, and informant switch ensures they cannot be.

---

References

Primary sources cited

Agnew-Blais, J. C. et al. (2016). "Evaluation of the Persistence, Remission, and Emergence of Attention-Deficit/Hyperactivity Disorder in Young Adulthood." JAMA Psychiatry, 73(7), 713–720. DOI: 10.1001/jamapsychiatry.2016.0465

Caye, A. et al. (2016). "Attention-Deficit/Hyperactivity Disorder Trajectories From Childhood to Young Adulthood: Evidence From a Birth Cohort Supporting a Late-Onset Syndrome." JAMA Psychiatry, 73(7), 705–712. DOI: 10.1001/jamapsychiatry.2016.0383

Faraone, S. V. & Biederman, J. (2016). "Can Attention-Deficit/Hyperactivity Disorder Onset Occur in Adulthood?" JAMA Psychiatry, 73(7), 655–656. DOI: 10.1001/jamapsychiatry.2016.0400

Moffitt, T. E. et al. (2015). "Is Adult ADHD a Childhood-Onset Neurodevelopmental Disorder? Evidence From a Four-Decade Longitudinal Cohort Study." American Journal of Psychiatry, 172(10), 967–977. DOI: 10.1176/appi.ajp.2015.14101266, PMC4591104

Sibley, M. H. et al. (2018). "Late-Onset ADHD Reconsidered With Comprehensive Repeated Assessments Between Ages 10 and 25." American Journal of Psychiatry, 175(2), 140–149. DOI: 10.1176/appi.ajp.2017.17030298, PMC5814300

Sibley, M. H. et al. (2022). "Variable Patterns of Remission From ADHD in the Multimodal Treatment Study of ADHD." American Journal of Psychiatry, 179(2), 142–151. DOI: 10.1176/appi.ajp.2021.21010032, PMC8810708

Staley, B. et al. (2024). "Attention-Deficit/Hyperactivity Disorder Diagnosis, Treatment, and Telehealth Use in Adults — National Center for Health Statistics Rapid Surveys System, United States, October–November 2023." Morbidity and Mortality Weekly Report, 73(40), 890–895. DOI: 10.15585/mmwr.mm7340a1

---

Appendix: Bind table for executive summary claims

Claim	Source	Identity
Dunedin childhood ADHD n=61 (6.0%); adult ADHD at age 38 n=31 (3.1%)	Moffitt et al. 2015	Am J Psychiatry 172:967–977, DOI 10.1176/appi.ajp.2015.14101266, PMC4591104
90% of age-38 adult-ADHD cases lacked prospectively diagnosed childhood ADHD (28 of 31)	Moffitt et al. 2015	Am J Psychiatry 172:967–977, DOI 10.1176/appi.ajp.2015.14101266, PMC4591104
5% of childhood cases still met adult criteria (3 of 61)	Moffitt et al. 2015	Am J Psychiatry 172:967–977, DOI 10.1176/appi.ajp.2015.14101266, PMC4591104
Pelotas: childhood (age 11) n=393 (8.9%); young-adult (age 18-19) n=492 (12.2%)	Caye et al. 2016	JAMA Psychiatry 73:705–712, DOI 10.1001/jamapsychiatry.2016.0383
Pelotas: 60/393 (17.2%) childhood cases persisted; 60/492 (12.2%) YA had C-ADHD; 416/492 (84.6%) did not	Caye et al. 2016	JAMA Psychiatry 73:705–712, DOI 10.1001/jamapsychiatry.2016.0383
Pelotas childhood SDQ screen sensitivity 85.7%, specificity 67.4%	Caye et al. 2016	JAMA Psychiatry 73:705–712, DOI 10.1001/jamapsychiatry.2016.0383
E-Risk: childhood n=247 (12.1% of 2040); age-18 n=166 (8.1%)	Agnew-Blais et al. 2016	JAMA Psychiatry 73:713–720, DOI 10.1001/jamapsychiatry.2016.0465
E-Risk: 54/247 (21.9%) childhood cases persisted; 112/166 (67.5%) age-18 cases no childhood dx	Agnew-Blais et al. 2016	JAMA Psychiatry 73:713–720, DOI 10.1001/jamapsychiatry.2016.0465
Informant switch, false-positive paradox, subthreshold childhood symptoms commentary	Faraone & Biederman 2016	JAMA Psychiatry 73:655–656, DOI 10.1001/jamapsychiatry.2016.0400
MTA LNCG: 289 recruited; 31 baseline ADHD out; 19 insufficient follow-up; N=239 for analysis	Sibley et al. 2018	Am J Psychiatry 175:140–149, DOI 10.1176/appi.ajp.2017.17030298, PMC5814300
Adolescent screens 97/239 (40.6%) → 6 (2.5%) cross-situational; 93.8% excluded	Sibley et al. 2018	Am J Psychiatry 175:140–149, DOI 10.1176/appi.ajp.2017.17030298, PMC5814300
Adult screens 47/239 (19.7%) → 2 (0.8%) cross-situational; 95.7% excluded	Sibley et al. 2018	Am J Psychiatry 175:140–149, DOI 10.1176/appi.ajp.2017.17030298, PMC5814300
Substance exclusions: 3 of 21 adolescent first-impairment cases; 14 of 24 adult first-impairment cases	Sibley et al. 2018	Am J Psychiatry 175:140–149, DOI 10.1176/appi.ajp.2017.17030298, PMC5814300
6 adolescent-onset: 4 adolescence-limited; 2 late-identified (above-average childhood symptoms)	Sibley et al. 2018	Am J Psychiatry 175:140–149, DOI 10.1176/appi.ajp.2017.17030298, PMC5814300
2 adult-onset: each one wave post-onset, not independent of complex psychiatric history	Sibley et al. 2018	Am J Psychiatry 175:140–149, DOI 10.1176/appi.ajp.2017.17030298, PMC5814300
MTA Combined Type: recovery 9.1% (51); fluctuating 63.8% (356); stable persistence 10.8% (60); stable partial remission 15.6% (87)	Sibley et al. 2022	Am J Psychiatry 179:142–151, DOI 10.1176/appi.ajp.2021.21010032, PMC8810708
US adults Oct-Nov 2023 NCHS: current diagnosed ADHD 6.0% (5.3–6.8); 55.9% first diagnosed ≥18; 36.5% untreated; 33.4% stimulants	Staley et al. 2024	MMWR 73(40):890–895, DOI 10.15585/mmwr.mm7340a1