Treatments for ADHD in Children and Adolescents: A Systematic Review

The review aims were developed in consultation with the Agency for Healthcare Research and Quality (AHRQ), the Patient-Centered Outcomes Research Institute, the topic nominator American Academy of Pediatrics (AAP), key informants, a technical expert panel (TEP), and public input. The TEP reviewed the protocol and advised on key outcomes. Subgroup analyses and key outcomes were prespecified. The review is registered in PROSPERO (#CRD42022312656) and the protocol is available on the AHRQ Web site as part of a larger evidence report on ADHD. The systematic review followed Methods of the (AHRQ) Evidence-based Practice Center Program.⁸ 

Population: Children or adolescents with a clinical diagnosis of ADHD, age <18 years

Interventions: Any ADHD treatment, alone or in combination, and ≥4 weeks’ treatment

Comparators: No treatment, waitlist, placebo, passive comparators, or active comparators

Outcomes: Patient health and psychosocial outcomes

Setting: Any

Study designs: RCTs for medication; RCTs, controlled clinical trials without random assignment, or cohort studies comparing 1 or more treatment groups for nondrug treatments. Studies either had to be large or demonstrate that they could detect effects as a standalone study (operationalized as ≥100 participants or a power calculation)

Other limiters: English-language (to ensure transparency for a US guideline), published from 1980

We searched the databases PubMed, Embase, PsycINFO, ERIC, and ClinicalTrials.gov. We identified reviews for reference-mining through PubMed, Cochrane Database of Systematic Reviews, Campbell Collaboration, What Works in Education, PROSPERO, ECRI Guidelines Trust, G-I-N, and ClinicalKey. The search underwent peer review; the full strategy is in the Online Appendix. All citations were reviewed by trained literature reviewers supported by machine learning to ensure no studies were inadvertently missed. Two independent reviewers assessed full-text studies for eligibility. Publications reporting on the same participants were consolidated into 1 record so that no study entered the analyses more than once. The TEP reviewed studies to ensure all were captured.

The data abstraction form included extensive guidance to aid reproducibility and standardization in recording study details, outcomes,^9 –12  study quality, and applicability. One reviewer abstracted data, and a methodologist checked its accuracy and completeness. Data are publicly available in the Systematic Review Data Repository.

We assessed 6 domains¹³ : Selection, performance, attrition, detection, reporting, and study-specific biases (Supplemental Figs 6 and 7).

We organized analyses by treatment and comparison type. We grouped treatments according to intervention content and target (eg, youth or parents). The intervention taxonomy differentiated medication, psychosocial interventions, parent support, nutrition and supplements, neurofeedback, neurostimulation, physical exercise, complementary medicine, school interventions, and provider approaches. We differentiated effects versus passive control groups (eg, placebo) and comparative effects (ie, comparing to an alternative treatment). The following outcomes were selected as key outcomes: (1) ADHD symptoms (eg, ADHD Rating Scale^14,15 ), (2) disruptive behavior (eg, conduct problems), (3) broadband measures (eg, Clinical Global Impression¹⁶ ), (4) functional impairment (eg, Weiss Functional Impairment Rating Scale^17,18 ), (5) academic performance (eg, grade point average), (6) appetite suppression, and (7) number of participants reporting adverse events.

Studies reported on a large range of outcome measures as documented in the evidence table in the Online Appendix. To facilitate comparisons across studies, we converted outcomes to scale-independent standardized mean differences (SMDs) for continuous symptom outcome variables and relative risks (RRs) for categorical reports, presenting summary estimates and 95% confidence intervals (CIs) for all analyses. We used random-effects models performed in R with Metafor_v4.2-0 for statistical pooling, correcting for small numbers of studies when necessary, to synthesize available evidence.¹⁹  We conducted sensitivity analyses for all analyses that included studies without random assignment. We also compared treatment effectiveness indirectly across studies in meta-regressions that added potential, prespecified effect modifiers to the meta-analytic model. In particular, we assessed whether ADHD presentation or cooccurring disorders modified intervention effects. We tested for heterogeneity using graphical displays, documented I² statistics (values >50% are highlighted in the text), and explored sources of heterogeneity in subgroup and sensitivity analyses.²⁰ 

We assessed publication bias with Begg and Egger tests^21,22  and used the trim-and-fill methods for alternative estimates where necessary.²³  Applicability of findings to real-world clinical practices in typical US settings was assessed qualitatively using AHRQ’s Methods Guide. An overall strength of evidence (SoE) assessment communicating our confidence in each finding was determined initially by 1 researcher with experience in use of specified standardized criteria²⁴  (Supplemental Information), then discussed with the study team. We downgraded SoE for study limitations, imprecision, inconsistency, and reporting bias, and we differentiated high, moderate, low, and insufficient SoE.

We screened 23 139 citations and retrieved 7534 publications as full text against the eligibility criteria. In total, 312 treatment studies, reported in 540 publications (see list of included studies in the Online Appendix), met eligibility criteria (Fig 1).

Although studies from 1980 were eligible, the earliest study meeting all eligibility criteria was from 1995. All included studies are documented in the evidence table in the Supplemental Information. The following highlights key findings. Results for intervention groups and individual studies, subgroup and sensitivity analyses, characteristics of participants and interventions contributing to the analyses, and considerations that determined the SoE for results are documented in the Online Appendix.

As a class, traditional stimulants (methylphenidate, amphetamines) significantly improved ADHD symptom severity (SMD, −0.88; CI, −1.13 to −0.63; studies = 12; n = 1620) and broadband measures (RR, 0.38; CI, 0.30–0.48; studies = 12; n = 1582) (both high SoE), but not functional impairment (SMD, 1.00; CI, −0.25 to 2.26; studies = 4; n = 540) (Fig 2, Supplemental Fig 8, Supplemental Table 1). Methylphenidate formulations significantly improved ADHD symptoms (SMD, −0.68; CI, −0.91 to −0.46; studies = 7; n = 863) (Fig 2, Supplemental Table 1) and broadband measures (SMD, 0.66; CI, 0.04–1.28; studies = 2; n = 302). Only 1 study assessed academic performance, reporting large improvements compared with a control group (SMD, −1.37; CI, −1.72 to −1.03; n = 156) (Supplemental Fig 9).²⁵  Methylphenidate statistically significantly suppressed appetite (RR, 2.80; CI, 1.47–5.32; studies = 8; n = 1110) (Fig 3), and more patients reported adverse events (RR, 1.32; CI, 1.25–1.40; studies = 6; n = 945). Amphetamine formulations significantly improved ADHD symptoms(SMD, −1.16; CI, −1.64 to −0.67; studies = 5; n = 757) (Fig 2, Supplemental Table 1) but not broadband measures(SMD, 0.68; CI, −0.72 to 2.08; studies = 3; n = 561) (Supplemental Fig 9). Amphetamines significantly suppressed appetite (RR, 7.08; CI, 2.72–18.42; studies = 8; n = 1229) (Fig 3), and more patients reported adverse events (RR, 1.41; CI, 1.25–1.58; studies = 8; n = 1151). Modafinil (US Food and Drug Administration [FDA]-approved to treat narcolepsy and sleep apnea but not ADHD) in each individual study significantly improved ADHD symptoms, but aggregated estimates were nonsignificant (SMD, −0.76; CI, −1.75 to 0.23; studies = 4; n = 667) (Fig 2, Supplemental Table 1) because of high heterogeneity (I² = 91%). It did not improve broadband measures (RR, 0.49; CI, −0.12 to 2.07; studies = 3; n = 539) (Supplemental Fig 9), and it significantly suppressed appetite (RR, 4.44; CI, 2.27–8.69; studies = 5; n = 780) (Fig 3).

As a class, nonstimulants significantly improved ADHD symptoms (SMD, −0.52; CI, −0.59 to −0.46; studies = 37; n = 6065; high SoE) (Fig 2, Supplemental Table 1), broadband measures (RR, 0.66; CI, 0.58–0.76; studies = 12; n = 2312) (Supplemental Fig 8), and disruptive behaviors (SMD, 0.66; CI, 0.22–1.10; studies = 4; n = 523), but not functional impairment (SMD, 0.20; CI, −0.05 to 0.44; studies = 6; n = 1163). Norepinephrine reuptake inhibitors (NRI) improved ADHD symptoms (SMD, −0.55; CI, −0.62 to −0.47; studies=28; n = 4493) (Fig 2, Supplemental Table 1) but suppressed appetite (RR, 3.23; CI, 2.40–4.34; studies = 27; n = 4176) (Fig 3), and more patients reported adverse events (RR, 1.31; CI, 1.18–1.46; studies = 15; n = 2600). Alpha-agonists (guanfacine and clonidine) improved ADHD symptoms (SMD, −0.52; CI, −0.67 to −0.37; studies = 11; n = 1885) (Fig 2, Supplemental Table 1), without (guanfacine) significantlysuppressing appetite (RR, 1.49; CI, 0.94–2.37; studies = 4; n = 919) (Fig 3), but more patients reported adverse events (RR, 1.21; CI, 1.11–1.31; studies = 14, n = 2544).

One study compared amphetamine versus methylphenidate, head-to-head, finding more improvement in ADHD symptoms (SMD, −0.46; CI, −0.73 to −0.19; n = 222) and broadband measures (SMD, 0.29; CI, 0.02–0.56; n = 211), but not functional impairment (SMD, 0.16; CI, −0.11 to 0.43; n = 211),²⁶  with lisdexamfetamine (an amphetamine) than osmotic-release oral system methylphenidate. No difference was found in appetite suppression(RR, 1.01; CI, 0.72–1.42; studies = 2, n = 414) (Fig 3) or adverse events (RR, 1.11; CI, 0.93–1.33; study = 1, n = 222). Indirect comparisons yielded significantly larger effects for amphetamine than methylphenidate in improving ADHD symptoms (P = .02) but not broadband measures (P = .97) or functional impairment (P = .68). Stimulants did not differ in appetite suppression (P = .08) or adverse events (P = .35).

One study provided information on NRI versus alpha-agonists by directly comparing an alpha-agonist (guanfacine) with an NRI (atomoxetine),²⁷  finding significantly greater improvement in ADHD symptoms with guanfacine (SMD, −0.47; CI, −0.73 to −0.2; n = 226) but not a broadband measure (RR, 0.84; CI, 0.68–1.04; n = 226). It reported less appetite suppression for guanfacine (RR, 0.48; CI, 0.27–0.83; n = 226) but no difference in adverse events (RR, 1.14; CI, 0.97–1.34; n = 226). Indirect comparisons did not indicate significantly different effect sizes for ADHD symptoms (P = .90), disruptive behaviors (P = .31), broadband measures (P = .41), functional impairment (P = .46), or adverse events (P = .06), but suggested NRIs more often suppressed appetite compared with guanfacine (P = .01).

Studies directly comparing nonstimulants versus stimulants (all were the NRI atomoxetine and stimulants methylphenidate in all but 1) tended to favor stimulants but did not yield significance for ADHD symptom severity (SMD, 0.23; CI, −0.03 to 0.49; studies = 7; n = 1611) (Fig 2). Atomoxetine slightly but statistically significantly produced greater improvements in disruptive behaviors (SMD, −0.08; CI, −0.14 to −0.03; studies = 4; n = 608) (Supplemental Fig 10) but not broadband measures (SMD, −0.16; CI, −0.36 to 0.04; studies = 4; n = 1080) (Supplemental Fig 9). They did not differ significantly in appetite suppression (RR, 0.82; CI, 0.53–1.26; studies = 8; n = 1463) (Fig 3) or number with adverse events (RR, 1.11; CI, 0.90–1.37; studies = 4; n = 756). Indirect comparisons indicated significant differences favoring stimulants over nonstimulants in improving ADHD symptom severity (P < .0001), broadband measures (P = .0002), and functional impairment (P = .04), but not appetite suppression (P = .31) or number with adverse events (P = .12).

Several studies assessed whether adding nonstimulant to stimulant medication (all were alpha-agonists added to different stimulants) improved outcomes compared with stimulant medication alone, yielding a small but significant additional improvement in ADHD symptoms (SMD, −0.36; CI, −0.52 to −0.19; studies = 5; n = 724) (Fig 4).

We identified 32 studies evaluating psychosocial, psychological, or behavioral interventions targeting ADHD youth, either alone or combined with components for parents and teachers. Interventions were highly diverse, and most were complex with multiple components (see supplemental results in the Online Appendix). They significantly improved ADHD symptoms (SMD, −0.35; CI, −0.51 to −0.19; studies = 14; n = 1686; moderate SoE) (Fig 4), even when restricting to RCTs only (SMD, −0.36; CI, −0.53 to −0.19; removing high-risk-of-bias studies left 7 with similar effects SMD, −0.38; CI, −0.69 to −0.07), with minimal heterogeneity (I² = 52%); but not disruptive behaviors (SMD, −0.18; CI, −0.48 to 0.12; studies = 8; n = 947) or academic performance (SMD, −0.07; CI, −0.49 to 0.62; studies = 3; n = 459) (Supplemental Fig 11).

We identified 19 studies primarily targeting parents of youth aged 3 to 18 years, though only 3 included teenagers. Interventions were highly diverse (see Online Appendix), but significantly improved ADHD symptoms (SMD, −0.31; CI, −0.57 to −0.05; studies = 11; n = 1078; low SoE) (Fig 4), even when restricting to RCTs only (SMD, −0.35; CI, −0.61 to −0.09; removing high-risk-of-bias studies yielded the same point estimate, but CIs were wider, and the effect was nonsignificant SMD, −0.31; CI, −0.76 to 0.14). There was some evidence of publication bias (Begg P = .16; Egger P = .02), but the trim and fill method to correct it found a similar effect (SMD, −0.43; CI, −0.63 to −0.22). Interventions improved broadband scores (SMD, 0.41; CI, 0.23–0.58; studies = 7; n = 613) and disruptive behaviors (SMD, −0.52; CI, −0.85 to −0.18; studies = 4; n = 357) but not functional impairment (SMD, 0.35; CI, −0.69 to 1.39; studies = 3; n = 252) (all low SoE) (Supplemental Fig 12).

We identified 10 studies, mostly for elementary or middle schools (see Online Appendix). Interventions did not significantly improve ADHD symptoms (SMD, −0.50; CI, −1.05 to 0.06; studies = 5; n = 822; moderate SoE) (Fig 4), but there was evidence of heterogeneity (I² = 87%). Although most studies reported improved academic performance, this was not statistically significant across studies (SMD, −0.19; CI, −0.48 to 0.09; studies = 5; n = 854) (Supplemental Fig 13).

We identified 22 studies, for youth aged 6 to 17 years without intellectual disability (see Online Appendix). Cognitive training did improve ADHD symptoms (SMD, −0.37; CI, −0.65 to −0.06; studies = 12; n = 655; low SoE) (Fig 4), with some heterogeneity (I² = 65%), but not functional impairment (SMD, 0.41; CI, −0.24 to 1.06; studies = 5; n = 387) (Supplemental Fig 14) or disruptive behaviors (SMD, −0.29; CI, −0.84 to 0.27; studies [all RCTs] = 5; n = 337). It improved broadband measures (SMD, 0.50; CI, 0.12–0.88; studies = 6; n = 344; RCTs only: SMD, 0.43; CI, −0.06 to 0.93) (both low SoE). It did not increase adverse events (RR, 3.30; CI, 0.03–431.32; studies = 2; n = 402).

We identified 21 studies: Two-thirds involved θ/β EEG marker modulation, and one-third modulation of slow cortical potentials (see Online Appendix). Neurofeedback significantly improved ADHD symptoms (SMD, −0.44; CI, −0.65 to −0.22; studies = 12; n = 945; low SoE) (Fig 4), with little heterogeneity (I² = 33%); restricting to the 10 RCTs yielded the same point estimate, also statistically significant (SMD, −0.44; CI, −0.71 to −0.16). Neurofeedback did not systematically improve disruptive behaviors (SMD, −0.33; CI, −1.33 to 0.66; studies = 4; n = 372), or functional impairment (SMD, 0.21; CI, −0.14 to 0.55; studies = 3; n = 332) (Supplemental Fig 15).

We identified 39 studies with highly diverse nutrition interventions (see Online Appendix), including omega-3 (studies = 13), vitamins (studies = 3), or diets (studies = 3), and several evaluated supplements as augmentation to stimulants. Most were placebo-controlled. Across studies, interventions improved ADHD symptoms (SMD, −0.39; CI, −0.67 to −0.12; studies = 23; n = 2357) (Fig 4), even when restricting to RCTs (SMD, −0.32; CI, −0.55 to −0.08), with high heterogeneity (I² = 89%) but no publication bias. The group of nutritional approaches also improved disruptive behaviors (SMD, −0.28; CI, −0.37 to −0.18; studies [all RCTs] = 5; n = 360) (Supplemental Fig 16, low SoE), without increasing the number reporting adverse events (RR, 0.77; CI, 0.47–1.27; studies = 8; n = 735). However, we did not identify any specific supplements that consistently improved outcomes, including omega-3 (eg, ADHD symptoms: SMD, −0.11; CI, −0.45, 0.24; studies = 7; n = 719; broadband measures: SMD, 0.04; CI, −0.24 to 0.32; studies = 7; n = 755, low SoE).

We identified 6 studies assessing acupuncture, homeopathy, and hippotherapy. They did not individually or as a group significantly improve ADHD symptoms (SMD, −0.15; CI, −1.84 to 1.53; studies = 3; n = 313) (Fig 4) or improve other outcomes across studies (eg, broadband measures: SMD, 0.03; CI, −3.66 to 3.73; studies = 2; n = 218) (Supplemental Fig 17).

Eleven identified studies evaluated a combination of medication- and youth-directed psychosocial treatments. Most allowed children to have common cooccurring conditions, but intellectual disability and severe neurodevelopmental conditions were exclusionary. Medication treatments were stimulant or atomoxetine. Psychosocial treatments included multimodal psychosocial treatment, cognitive behavioral therapy, solution-focused therapy, behavioral therapy, and a humanistic intervention. Studies mostly compared combinations of medication and psychosocial treatment to medication alone, rather than no treatment or placebo. Combined therapy did not statistically significantly improve ADHD symptoms across studies (SMD, −0.36; CI, −0.73 to 0.01; studies = 7; n = 841; low SoE; only 2 individual studies reported statistically significant effects) (Fig 5) or broadband measures (SMD, 0.42; CI, −0.72 to 1.56; studies = 3; n = 171), but there was indication of heterogeneity (I² = 71% and 62%, respectively).

We found little evidence that either ADHD presentation (inattentive, hyperactive, combined-type) or cooccurring psychiatric disorders modified treatment effects on any ADHD outcome, but few studies addressed this question systematically (see Online Appendix).

Only a very small number of studies (33 of 312) reported on outcomes at or beyond 12 months of follow-up (see Online Appendix). Many did not report on key outcomes of this review. Studies evaluating combined psychosocial and medication interventions, such as the multimodal treatment of ADHD study,²⁸  did not find sustained effects beyond 12 months. Analyses for medication, psychosocial, neurofeedback, parent support, school intervention, and provider-focused interventions did not find sustained effects for more than a single study reporting on the same outcome. No complementary medicine, neurostimulation, physical exercise, or cognitive training studies reported long-term outcomes.

We identified a large body of evidence contributing to knowledge of ADHD treatments. A substantial number of treatments have been evaluated in strong study designs that provide evidence statements regarding the effects of the treatments on children and adolescents with ADHD. The body of evidence shows that numerous intervention classes significantly improve ADHD symptom severity. This includes large but variable effects for amphetamines, moderate-sized effects for methylphenidate, NRIs, and alpha-agonists, and small effects for youth-directed psychosocial treatment, parent support, neurofeedback, and cognitive training. The SoE for effects on ADHD symptoms was high across FDA-approved medications (methylphenidate, amphetamines, NRIs, alpha-agonists); moderate for psychosocial interventions; and low for parent support, neurofeedback, and nutritional interventions. Augmentation of stimulant medication with non-stimulants produced small but significant additional improvement in ADHD symptoms over stimulant medication alone (low SoE).

We also summarized evidence for other outcomes beyond specific ADHD symptoms and found that broadband measures (ie, global clinical measures not restricted to assessing specific symptoms and documenting overall psychosocial adjustment), methylphenidate (low SoE), nonstimulant medications (moderate SoE), and cognitive training (low SoE) yielded significant, medium-sized effects, and parent support small effects (moderate SoE). For disruptive behaviors, nonstimulant medications (high SoE) and parent support (low SoE) produced significant improvement with medium effect. No treatment modality significantly improved functional impairment or academic performance, though the latter was rarely assessed as a treatment outcome.

The enormous variability in treatment components and delivery of youth-directed psychotherapies, parent support, neurofeedback, and nutrition and supplement therapies, and in ADHD outcomes they have targeted, complicates the synthesis and meta-analysis of their effects compared with the much more uniform interventions, delivery, and outcome assessments for medication therapies. Moreover, most psychosocial and parent support studies compared an active treatment against wait list controls or treatment as usual, which did not control well for the effects of parent or therapist attention or other nonspecific effects of therapy, and they have rarely been able to blind adequately either participants or study assessors to treatment assignment.^29,30  These design limitations weaken the SoE for these interventions.

The large number of studies, combined with their medium-to-large effect sizes, indicate collectively and with high SoE that FDA-approved medications improve ADHD symptom severity, broadband measures, functional impairment, and disruptive behaviors. Indirect comparison showed larger effect sizes for stimulants than for nonstimulants in improving ADHD symptoms and functional impairment. Results for amphetamines and methylphenidate varied, and we did not identify head-to-head comparisons of NRIs versus alpha-agonists that met eligibility criteria. Despite compelling evidence for their effectiveness, stimulants and nonstimulants produced more adverse events than did other interventions, with a high SoE. Stimulants and nonstimulant NRIs produced significantly more appetite suppression than placebo, with similar effect sizes for methylphenidate, amphetamine, and NRI, and much larger effects for modafinil. Nonstimulant alpha-agonists (specifically, guanfacine) did not suppress appetite. Rates of other adverse events were similar between NRIs and alpha-agonists.

Perhaps contrary to common belief, we found no evidence that youth-directed psychosocial and medication interventions are systematically better in improving ADHD outcomes when delivered as combination treatments^31 –33 ; both were effective as monotherapies, but the combination did not signal additional statistically significant benefits (low SoE). However, it should be noted that few psychosocial and medication intervention combinations have been studied to date. We also found that treatment outcomes did not vary with ADHD presentation or the presence of cooccurring psychiatric disorders, but indirect analyses are limited in detecting these effect modifiers, and more research is needed. Furthermore, although children of all ages were eligible for inclusion in the review, we note that very few studies assessed treatments (especially medications) in children <6 years of age; evidence is primarily available for school-age children and adolescents. Finally, despite the research volume, we still know little about long-term effects of ADHD treatments. The limited available body of evidence suggests that most interventions, including combined medication and psychological treatment, yield few significant long-term improvements for most ADHD outcomes.

This review provides compelling evidence that numerous, diverse treatments are available and helpful for the treatment of ADHD. These include stimulant and nonstimulant medications, youth-targeted psychosocial treatments, parent support, neurofeedback, and cognitive training, though nonmedication interventions appear to have considerably weaker effects than medications on ADHD symptoms. Nonetheless, the body of evidence provides youth with ADHD, their parents, and health care providers with options.

The paucity of head-to-head studies comparing treatments precludes research-based recommendations regarding which is likely to be most helpful and which should be tried first, and decisions need to be based on clinical considerations and patient preferences. Stimulant and nonstimulant NRI medications, separately and in head-to-head comparisons, have shown similar effectiveness and rates of side effects, including appetite suppression, across identified studies. The moderate effect sizes for nonstimulant alpha-agonists, their low rate of appetite suppression, and their evidence for effectiveness in augmenting the effects of stimulant medications in reducing ADHD symptom severity provides additional treatment options. Furthermore, we found low SoE that neurofeedback and cognitive training improve ADHD symptoms. We also found that nutritional supplements and dietary interventions improve ADHD symptoms and disruptive behaviors. The SoE for nutritional interventions, however, is still low, and despite the research volume, we did not identify systematic benefits for specific supplements.

Clinical guidelines currently advise starting treatment of youth >6 years of age with FDA-approved medications,³³  which the findings of this review support. Furthermore, FDA-approved medications have been shown to significantly improve broadband measures, and nonstimulant medications have been shown to improve disruptive behaviors, suggesting their clinical benefits extend beyond improving only ADHD symptoms. Clinical guidelines for preschool children advise parent training and/or classroom behavioral interventions as the first line of treatment, if available. These recommendations remain supported by the present review, given the paucity of studies in preschool children in general, and because many existing studies, in particular medication and youth-directed psychosocial interventions, do not include young children.^31 –33 

This review incorporated publications dating from 1980, assessing diverse intervention targets (youth, parent, school) and ADHD outcomes across numerous functional domains. Limitations in its scope derive from eligibility criteria. Requiring treatment of 4 weeks ensured that interventions were intended as patient treatment rather than proof of concept experiments, but it also excluded some early studies contributing to the field and other brief but intense psychosocial interventions. Requiring studies to be sufficiently large to detect effects excluded smaller studies that contribute to the evidence base. We explicitly did not restrict to RCTs (ie, a traditional medical study design), but instead identified all studies with concurrent comparators so as not to bias against psychosocial research; nonetheless, the large majority of identified studies were RCTs. Our review aimed to provide an overview of the diverse treatment options and we abstracted findings regardless of the suitability of the study results for meta-analysis. Although many ADHD treatments are very different in nature and the clinical decision for 1 treatment approach over another is likely not made primarily on effect size estimates, future research could use the identified study pool and systematically analyze comparative effectiveness of functionally interchangeable treatments in a network meta-analysis, building on previous work on medication options.³⁴ 

Future studies of psychosocial, parent, school-based, neurofeedback, and nutritional treatments should employ more uniform interventions and study designs that provide a higher SoE for effectiveness, including active attention comparators and effective blinding of outcome assessments. Higher-quality studies are needed for exercise and neuromodulation interventions. More trials are needed that compare alternative interventions head-to-head or compare combination treatments with monotherapy. Clinical trials should assess patient-centered outcomes other than ADHD symptoms, including functional impairment and academic performance. Much more research is needed to assess long-term treatment effectiveness, compliance, and safety, including in preschool youth. Studies should assess patient characteristics as modifiers of treatment effects, to identify which treatments are most effective for which patients. To aid discovery and confirmation of these modifiers, studies should make publicly available all individual-level demographic, clinical, treatment, and outcome data.

We thank the following individuals providing expertise and helpful comments that contributed to the systematic review: Esther Lee, Becky Nguyen, Cynthia Ramirez, Erin Tokutomi, Ben Coughli, Jennifer Rivera, Coleman Schaefer, Cindy Pham, Jerusalem Belay, Anne Onyekwuluje, Mario Gastelum, Karin Celosse, Samantha Fleck, Janice Kang, and Sreya Molakalaplli for help with data acquisition. We thank Kymika Okechukwu, Lauren Pilcher, Joanna King, and Robyn Wheatley from the American Academy of Pediatrics; Jennie Dalton and Paula Eguino Medina from the Patient-Centered Outcomes Research Institute; Christine Chang and Kim Wittenberg from AHRQ; and Mary Butler from the Minnesota Evidence-based Practice Center. We thank Glendy Burnett, Eugenia Chan, MD, MPH; Matthew J. Gormley, PhD; Laurence Greenhill, MD; Joseph Hagan, Jr, MD; Cecil Reynolds, PhD; Le’Ann Solmonson, PhD, LPC-S, CSC; and Peter Ziemkowski, MD, FAAFP; who served as key informants. We thank Angelika Claussen, PhD; Alysa Doyle, PhD; Tiffany Farchione, MD; Matthew J. Gormley, PhD; Laurence Greenhill, MD; Jeffrey M. Halperin, PhD; Marisa Perez-Martin, MS, LMFT; Russell Schachar, MD; Le’Ann Solmonson, PhD, LPC-S, CSC; and James Swanson, PhD; who served as a technical expert panel. Finally, we thank Joel Nigg, PhD; and Peter S. Jensen, MD; for their peer review of the data.

ADHD

attention-deficit/hyperactivity disorder

AHRQ

Agency for Healthcare Research and Quality

FDA

US Food and Drug Administration

CI

confidence interval

NRI

norepinephrine reuptake inhibitors

RCT

randomized controlled trial

RR

relative risk

SMD

standardized mean difference

SoE

strength of evidence

TEP

technical expert panel