Tricyclic antidepressants for attention deficit hyperactivity disorder (ADHD) in children and adolescents

Abstract

Background

Attention deficit hyperactivity disorder (ADHD) is a chronic neurodevelopmental disorder of childhood onset, which may persist into adulthood. ADHD has a significant impact on a child's daily life, affecting relationships and academic performance. Its core symptoms include developmentally inappropriate levels of inattention, hyperactivity, and impulsive behaviour. Tricyclic antidepressants (TCAs) are sometimes used as second line of treatment in the reduction of ADHD symptoms in children and adolescents with ADHD. However, their efficacy is not yet known.

Objectives

To assess the efficacy of TCAs in the reduction of ADHD symptoms within the broad categories of hyperactivity, impulsivity, and inattentiveness in young people aged 6 to 18 years with established diagnoses of ADHD.

Search methods

On 26 September 2013, we searched CENTRAL, Ovid MEDLINE, Embase, PsycINFO, CINAHL, seven other databases, and two trials registers. We also searched the reference lists of relevant articles, and contacted manufacturers and known experts in the field to determine if there were any ongoing trials or unpublished studies available.

Selection criteria

Randomised controlled trials (RCTs), including both parallel group and cross‐over study designs, of any dose of TCA compared with placebo or active medication in children or adolescents with ADHD, including those with comorbid conditions. 

Data collection and analysis

Working in pairs, three review authors independently screened records, extracted data, and assessed trial quality. We calculated the standardised mean differences (SMD) for continuous data, the odds ratio (OR) for dichotomous data, and 95% confidence intervals (CIs) for both. We conducted the meta‐analyses using a random‐effects model throughout. We used the Cochrane 'Risk of bias' tool to assess the risk of bias of each included trial and the GRADE approach to assess the quality of the body evidence.

Main results

We included six RCTs with a total of 216 participants. Five of the six trials compared desipramine with placebo; the remaining trial compared nortriptyline with placebo. One trial compared desipramine with clonidine and placebo, and another compared two TCAs (desipramine and clomipramine) with methylphenidate and placebo. Of the six trials, one RCT primarily assessed the efficacy of TCA in children with ADHD and comorbid tic or Tourette disorder, and another one trial was in children with comorbid tic disorder. RCTs that met our inclusion criteria varied both in design and quality, and none were free of bias. The quality of the evidence was low to very low according to our GRADE assessments.

TCA outperformed placebo regarding the proportions of patients achieving a predefined improvement of core ADHD symptom severity (OR 18.50, 95% CI 6.29 to 54.39, 3 trials, 125 participants, low quality evidence). In particular, there was evidence that desipramine improved the core symptoms of ADHD in children and adolescents as assessed by parents (SMD ‐1.42, 95% CI ‐1.99 to ‐0.85, 2 trials, 99 participants, low quality evidence), teachers (SMD ‐0.97, 95% CI ‐1.66 to ‐0.28, 2 trials, 89 participants, low quality evidence), and clinicians (OR 26.41, 95% CI 7.41 to 94.18, 2 trials, 103 participants, low quality evidence). Nortriptryline was also efficacious in improving the core symptoms of ADHD in children and adolescents as assessed by clinicians (OR 7.88, 95% CI 1.10 to 56.12). Desipramine and placebo were similar on "all‐cause treatment discontinuation" (RD ‐0.10, 95% CI ‐0.25 to 0.04, 3 trials, 134 participants, very low quality evidence). Desipramine appeared more efficacious than clonidine in reducing ADHD symptoms as rated by parents (SMD ‐0.90, 95% CI ‐1.40 to ‐0.40, 1 trial, 68 participants, very low quality evidence) in participants with ADHD and comorbid tics or Tourette syndrome.

Although this Cochrane Review did not identify serious adverse effects in patients taking TCAs, it did identify mild increases in diastolic blood pressure and pulse rates. Also, patients treated with desipramine had significantly higher rates of appetite suppression compared to placebo whilst nortriptyline resulted in weight gain. Other reported adverse effects included headache, confusion, sedation, tiredness, blurred vision, diaphoresis, dry mouth, abdominal discomfort, constipation, and urinary retention.

Authors' conclusions

Most evidence on TCAs relates to desipramine. Findings suggest that, in the short term, desipramine improves the core symptoms of ADHD, but its effect on the cardiovascular system remains an important clinical concern. Thus, evidence supporting the clinical use of desipramine for the treatment of children with ADHD is low.

Plain language summary

Tricyclic antidepressants for attention deficit hyperactivity disorder (ADHD) in children and adolescents

Background

ADHD affects a large number of children and adolescents, giving rise to problems with inattention, hyperactivity, and impulsivity. A variety of medications, including tricyclic antidepressants (TCAs), may be used to treat the core symptoms of ADHD.

Review question

We reviewed the evidence about the effectiveness of TCAs in improving the core symptoms of ADHD in children and adolescents with ADHD, and whether or not they might cause harm.

Trial characteristics

The evidence is current to 26 September 2013. Only six trials including 216 participants met our inclusion criteria. All trials were conducted in the USA and lasted between two and six weeks. Two trials were funded, in part, by the drug manufacturers.

Key results

Findings showed that TCAs, particularly desipramine, had a beneficial effect for children and adolescents with ADHD in terms of improving their core symptoms in the short‐term. However, TCAs also had unwanted cardiac effects, which may limit their use.

Quality of the evidence

The total number of included trials was small and each used many different outcome measures, making it difficult to combine their results. Further research is needed to test whether the findings from these trials are universally applicable; these should be conducted in different locations, with ethnically diverse participants, and use validated outcome measures to assess the core symptoms of ADHD.

Summary of findings

Summary of findings for the main comparison. Tricyclic antidepressants (TCAs) compared to placebo for attention deficit hyperactivity disorder (ADHD) in children and adolescents.

TCA compared to placebo for ADHD in children and adolescents
Patient or population: Children and adolescents with ADHD Settings: Outpatient clinic Intervention: TCA Comparison: Placebo
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	Number of participants (trials)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Placebo	TCA
Efficacy for ADHD symptoms (parents using a rating scale) ‐ desipramine	The figures for the control group could not be summarized in a single mean figure because the trials included in this analysis used different scales	The mean efficacy for ADHD symptoms (parents using a rating scale ) ‐ desipramine in the intervention groups was 1.42 standard deviations lower (1.99 to 0.85 lower)	‐	99 (2 trials)	⊕⊕⊝⊝ low1,2	SMD ‐1.42 (‐1.99 to ‐0.85)
Efficacy for ADHD symptoms (teachers using a rating scale) ‐ desipramine	The figures could not be summarized in a single mean figure because the trials included in this analysis used different scales	The mean efficacy for ADHD symptoms (teachers using a rating scale) ‐ desipramine in the intervention groups was 0.97 standard deviations lower (1.66 to 0.28 lower)	‐	89 (2 trials)	⊕⊕⊝⊝ low1,2	SMD ‐0.97 (‐1.66 to ‐0.28)
CGI (response rate) ‐ desipramine and nortriptyline	Trial population		OR 18.5 (6.29 to 54.39)	125 (3 trials)	⊕⊕⊝⊝ low2,3	‐
	83 per 1000	626 per 1000 (363 to 831)
	Medium risk population
	97 per 1000	665 per 1000 (403 to 854)
Clinical global impression (response rate) ‐ desipramine	Trial population		OR 26.41 (7.41 to 94.18)	103 (2 trials)	⊕⊕⊝⊝ low1,2	‐
	59 per 1000	623 per 1000 (317 to 855)
	Medium risk population
	48 per 1000	571 per 1000 (272 to 826)
All‐cause treatment discontinuation ‐ desipramine	Trial population		See comment	134 (3 trials)	⊕⊝⊝⊝ very low2,3,4	Risks were calculated from pooled risk differences
	159 per 1000	57 per 1000 (‐91 to 199)
	Medium risk population
	100 per 1000	36 per 1000 (‐57 to 125)
*The basis for the assumed risk (e.g. the median control group risk across trials) is provided in footnotes. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RR: Risk ratio; OR: Odds ratio; CGI: Clinical Global Impression.
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.

¹ Only 2 small RCTs could be combined. Therefore, a single trial could change the effect size calculated. For this reason the evidence was downgraded by 2.
 ² The statistical power to detect publication bias for this comparison in this review is low.
 ³ Only 3 small RCTs could be combined. Therefore, a single study could change the effect size calculated. For this reason the evidence was downgraded by 2.
 ⁴ Barely statistically significant heterogeneity was found. For this reason evidence was downgraded by 1

Summary of findings 2. Desipramine compared to clonidine for ADHD in children and adolescents.

Desipramine compared to clonidine for ADHD in children and adolescents
Patient or population: Children and adolescents with ADHD Settings: Outpatient clinic Intervention: Desipramine Comparison: Clonidine
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	Number of participants (trials)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Clonidine	Desipramine
Efficacy for ADHD symptoms (parent‐rated symptoms using VAS)	The mean (SD) endpoint score for the efficacy on ADHD symptoms assessed with the VAS in the clonidine group was 51.6 (28.2)	The mean efficacy for ADHD symptoms (parent‐rated symptoms using VAS) in the intervention groups was ‐0.90 lower (1.40 to 0.040 lower)	‐	68 (1 trial)	⊕⊝⊝⊝ very low1,2	SMD ‐90 (‐1.40 to ‐0.40)
*The basis for the assumed risk (e.g. the median control group risk across trials) is provided in footnotes. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval.
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.

¹ This trial had a cross‐over design and a carry‐over effect was found. For this reason the evidence was downgraded by 1.
 ² Only 1 small trial reported this outcome. Therefore, the risk that a single new study could change the effect calculated is very likely. For this reason the evidence was downgraded by 2.

Background

Description of the condition

Attention deficit hyperactivity disorder (ADHD) is the most common neuropsychiatric disorder of childhood. It affects 3% to 7% of school‐aged children in the United States and up to 1% of the population in the United Kingdom (WHO 1992; APA 2000), though prevalence rates vary according to the diagnostic criteria used (Jadad 1999; Polanczyk 2007). Using DSM‐IV criteria, for example, the prevalence amongst children and adolescents in Brazil was 5.8% to 18% (Rohde 1999; Guardiola 2000) and 8% amongst primary school‐aged children in Benin City, Nigeria (Ofovwe 2006).

The Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM‐IV; APA 2000) is published by the Task Force on Nomenclature and Statistics of the American Psychiatric Association. The manual covers all mental health disorders for both children and adults and provides detailed descriptions of categories of disorders as well as diagnostic criteria. It defines three ADHD clinical phenotypes ‐ inattentive, hyperactive‐impulsive, and combined. This classification is based on symptom count ‐ six for either inattentive or hyperactive‐impulsive and six in each category for combined. In describing the syndrome, theInternational Classification of Diseases (ICD‐10; WHO 1992) uses the specific diagnostic term of hyperkinetic disorder (HD), which comes close to meeting the criteria for the combined form of ADHD. For the purpose of this Cochrane Review, we will use the term ADHD to refer to both ADHD and hyperkinetic disorder.

ADHD is characterised by developmentally inappropriate symptoms of excessive inattentiveness, impulsivity, and hyperactivity. Its symptoms typically begin before the age of seven and can be disabling. Affected children are at greater risk for comorbid antisocial behaviour, poor academic or vocational performance, substance misuse, and other psychiatric disorders such as anxiety and depression (Wilens 2004; Biederman 2005). If left untreated, ADHD is associated with long‐term educational and social disadvantage (Swanson 1998; Harpin 2005).

The exact cause of ADHD is not known but multiple factors are thought to contribute to its aetiology. These include genetic, environmental, and neurobiological factors (Biederman 2005; Swanson 2007; Froehlich 2011). Dysfunction in catecholamine neurotransmission in the prefrontal cortex (PFC) has also been implicated (Arnsten 1996). The PFC is the part of the brain that is largely responsible for executive function. The PFC executive functions include inhibition, self‐monitoring, planning, working memory, and flexible shifting of actions. There is evidence for executive function impairment in people with ADHD (Pennington 1996; Barkley 1997). Also, Willcutt 2005 concluded that executive function appeared to be one of the important components of the complex neuropsychology of ADHD.

Description of the intervention

Symptoms of ADHD improve significantly with adequate treatment. Treatment aims to reduce core symptoms and improve functioning. This requires a well thought out and child‐specific treatment plan that describes methods and goals of treatment, means of monitoring care over time, and specific plans for follow‐up. It usually involves both medical and non‐medical (e.g. behavioural and psychosocial) interventions.

In choosing a medication treatment option, the National Institute of Clinical Excellence (NICE) recommends that the clinician consider the presence of co‐existing conditions; any potential side effects; specific issues around compliance, ability to take the medication, or both; any potential for drug misuse or diversion; and the preferences of the parents and patient.

Psychostimulants are generally considered to be the first‐line agents in the treatment of ADHD. However, atomoxetine ‐ a non‐stimulant medication ‐ is also being increasingly used as a first‐line treatment. Preliminary findings of a prospective, randomised open‐label trial by Kratochvil 2002, comparing atomoxetine with methylphenidate in the treatment of children with ADHD, showed therapeutic effects comparable to those of methylphenidate. However, in a community‐based randomised prospective study by Kemner 2005, atomoxetine was less effective in controlling ADHD symptoms than controlled‐release (OROS) methylphenidate.

Alpha‐agonists are also useful in the treatment of ADHD. The long acting alpha‐agonist, extended‐release clonidine, and the long acting alpha‐2A‐agonist, extended‐release guanfacine, have also been approved by the US Food and Drug Administration (FDA) for the treatment of ADHD as monotherapy or adjunctive treatment (in combination with a psychostimulant).

Some off‐label medications that have been used for the treatment of children and adolescents with ADHD include bupropion and TCAs. These agents are neither recommended by NICE nor approved by the FDA for the treatment of ADHD. However, they remain important alternatives or adjuncts in certain clinical situations, such as when a patient does not respond adequately to first‐line agents; occasions in which psychostimulant use may be less desirable, e.g. when there is serious concern about a comorbid illicit substance misuse problem; when intolerable side effects are encountered with first‐line agents such as severe exacerbation of tics with psychostimulants; and the presence of a medical condition that makes stimulant use difficult such as the presence of a pre‐existing cardiac condition. Occasionally, parents of children with ADHD may prefer not to use psychostimulants but accept alternative medication treatment.

The use of tricyclics as a class of medication in children and adolescents has generally been limited by concerns raised about their potential for cardiotoxicity and sudden death in this age‐group (Riddle 1991; Biederman 1995). Similar to other antidepressants, TCAs may increase the risk of suicide in some individuals (Hammad 2006; Schneeweiss 2010). These concerns have also limited their use in the treatment of ADHD. However, they have distinct advantages that make them a relevant alternative or adjunct in the treatment of ADHD, including a relatively long half‐life (approximately 12 hours), obviating the need to administer medications during school hours and lowering the potential for drug abuse or side effects, as well as their potentially positive effects on mood, anxiety, sleep, and tics.

How the intervention might work

TCAs appear to modulate various brain neurotransmitters, especially dopamine, norepinephrine, and serotonin, by blocking re‐uptake presynaptically, thus acting as agonists of these neurotransmitters. They increase the inhibitory influences of frontal cortical activity on subcortical structures through the dopaminergic or noradrenergic pathways, or both.

The noradrenergic system is broadly associated with the modulation of higher cortical functions, including attention, cognition, alertness, and executive function. Its activation greatly affects the performance of attention, especially in maintaining and increasing overall arousal level.

Extant literature from various sources, including genetic, neuropsychological, and structural and functional neuroimaging have shown that fronto‐striatal network abnormalities contribute to ADHD symptomatology (Bush 2005; Kelly 2007; Swanson 2007).

Pharmacological alterations in noradrenergic systems targeting symptoms of arousal or executive function have been shown to improve the core symptoms of ADHD. The efficacy of the TCAs in the treatment of ADHD is thought to be due to their actions on catecholamine reuptake, particularly noradrenaline (Biederman 1999). It is plausible that the effect of TCA is mediated by reducing arousal.

Why it is important to do this review

ADHD is a chronic condition, which if left untreated, has a detrimental effect on individuals and the community. Some people may not benefit from the use of stimulant medication and may have to rely on non‐stimulants such as TCAs. TCAs have a quicker onset of action than most other non‐stimulant medications (such as atomoxetine and clonidine) used in ADHD treatment. They are also readily available and relatively inexpensive. These medications, which are now off‐patent, have been in use in the paediatric population for several decades and are more likely to be used in countries where atomoxetine and stimulants are too expensive. It is therefore important to conduct a systematic review of TCAs for treatment of young people with ADHD to evaluate their benefits and safety.  

Objectives

To assess the efficacy of TCAs in the reduction of ADHD symptoms within the broad categories of hyperactivity, impulsivity, and inattentiveness in young people, aged six to 18 years, with established diagnoses of ADHD.

Methods

Criteria for considering studies for this review

Types of studies

Double‐blinded randomised controlled trials (RCTs), including both parallel group and cross‐over study designs.

Types of participants

Children and young people (six to 18 years of age) who have been diagnosed with ADHD using validated criteria from the DSM‐IV (APA 1994; APA 2000) or diagnostic categories from earlier systems (DSM‐III and DSM‐III‐R) (APA 1980; APA 1987), or who have an ICD‐10 diagnosis of hyperkinetic disorder (WHO 1992). We also included children and adolescents with comorbid conditions.

Types of interventions

1. TCAs (e.g. desipramine, imipramine, nortriptyline, clomipramine, and amitriptyline) versus placebo.

2. TCAs versus active medication (e.g. methylphenidate, atomoxetine, and clonidine).

3. TCAs versus TCAs.

For the primary analyses, we considered different TCAs as the same drug*. Trials of TCAs of any dose were included.

Types of outcome measures

Primary outcomes:

1. Severity of core ADHD symptoms as measured by validated clinician, teacher or parent report scales such as the Conners' Rating Scale (Conners 1994)*, Child Behaviour Checklist (CBCL; Achenbach 1991)*, or Clinical Global Impression (CGI; NIMH 1985)*.

2. The proportion of patients achieving a predefined improvement of core ADHD symptom severity as measured by validated clinician, teacher, or parent report scales such as the Conners' Rating Scale (Conners 1994)*, CBCL (Achenbach 1991)*, or CGI (NIMH 1985)*.

3. Treatment adverse effects such as cardiovascular symptoms (e.g. changes in heart rate or blood pressure), gastrointestinal symptoms, and changes in weight, appetite, or sleep pattern.

Secondary outcomes:

1. Severity of behavioural problems, e.g. oppositional defiant disorder or conduct disorder, measured by scales such as the CBCL (Achenbach 1991)* or CGI (NIMH 1985)*.

2. Severity of depressive symptoms measured by scales such as Children's Depression Inventory (Kovacs 1985)*.

3. Severity of anxiety symptoms measured by scales such as Revised Children's Manifest Anxiety Scale (RCMAS; Reynolds 1998).

4. Obsessive symptom severity as measured by scales such as Children's Yale‐Brown Obsessive Compulsive Scale (CYBOCS; Scahill 1997).

5. All‐cause treatment discontinuation*. The number of patients discontinuing the treatment assigned, irrespective of the reason for discontinuation.

We used the asterisked outcomes (*) to populate Table 1 and Table 2. We analysed three types of report scales ‐ clinician, teacher, or parent ‐ separately and jointly for the primary outcomes and treatment adverse effects.

Search methods for identification of studies

We devised a core search strategy in Ovid MEDLINE and adapted it for other databases using the appropriate syntax (see Appendix 1). We did not apply any date or language restrictions. Also, we did not limit our searches by publication status.

Electronic searches

We searched the following sources in June 2012 and again on 26 September 2013 (with the exception of the NIHR Archive):

1. Cochrane Central Register of Controlled Trials (CENTRAL), 2013 (Issue 9 of 12), part of The Cochrane Library.

2. Ovid MEDLINE (R) 1946 to September Week 3 2013.

3. Ovid MEDLINE(R) In‐Process & Other Non‐Indexed Citations, 25 September 2013.

4. Embase (Ovid) 1980 to 2013, Week 38.

5. PsycINFO (OVID), 1806 to September Week 4 2013.

6. CINAHLPlus (EBSCOhost) 1937 to current.

7. Science Citation Index Expanded (Web of Science) 1970 to 25 September 2013.

8. Social Sciences Citation Index (Web of Science) 1970 to 25 September 2013.

9. Conference Proceedings Citation Index ‐ Science (Web of Science), 1990 to 25 September 2013.

10. Conference Proceedings Citation Index ‐ Social Science & Humanities (Web of Science), 1990 to 25 September 2013.

11. WorldCat (OCLC) (worldcat.org/), limited to theses.

12. ICTRP (apps.who.int/trialsearch/AdvSearch.aspx).

13. ClinicalTrials.gov (clinicaltrials.gov/).

14. National Institute for Health Research Archive (nihr.ac.uk/Pages/NRRArchiveSearch.aspx), last searched 29 June 2012 (which is an archive and new records are not added).

Searching other resources

We also checked the reference lists of all relevant articles identified through the search strategy, and contacted the drug manufacturers of amitriptyline, clomipramine, desipramine, imipramine, and nortriptyline, and known experts in the field to determine whether there were any ongoing trials or unpublished results available.

Data collection and analysis

Selection of studies

Two review authors (JO, CS) assessed independently all trials identified by the search strategy for relevance. For articles deemed potentially suitable, or where more information was needed, the same review authors obtained the full‐text report and assessed it against the inclusion criteria. We resolved any disagreements as regards the selection of trials through consensus. If necessary, we involved a third review author (XC or UE) to settle any differences of opinion. Review authors were not blinded to the names of the trial authors or journals of publication. We kept records of reasons for exclusion (see 'Characteristics of excluded studies' section).  

Data extraction and management

All review authors independently extracted the following data from each included trial and entered the data into a pre‐designed table, which contained columns for details of participants, setting, methodology, interventions, and outcome data:

1. Trial procedures, including recruitment, diagnosis, medication, dosage, and duration.

2. Trial design (RCT or quasi‐RCT).

3. Randomisation method.

4. Method of allocation concealment.

5. Method of blinding.

6. Inclusion and exclusion criteria for participants.

7. Number of participants.

8. Gender of participants.

9. Age distribution.

10. Loss to follow‐up.

11. Premature discontinuation of trial and reasons for discontinuation.

12. Outcomes.

13. Methods of analysis (intention‐to‐treat (ITT), per protocol).

14. Comparability of groups at baseline.

One review author (XC) entered the data into Review Manager 2012 and a second author (JO) then checked them for accuracy. We resolved discrepancies through discussion and consensus. We only included outcome data that were measured using a scale or questionnaire that was reported as being valid and reliable in a peer‐reviewed journal.

Assessment of risk of bias in included studies

All review authors independently assessed risk of bias using criteria outlined in Chapter Eight of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We then compared the assessments for inconsistencies, and we resolved differences in interpretation by discussion and consensus. We assessed each included trial for bias across the following domains and assigned ratings of 'low', 'high', or 'unclear' risk of bias.

Sequence generation

Was the allocation sequence adequately generated? That is, the method used to generate the allocation sequence was described in sufficient detail to be able to assess whether it should have yielded comparable groups. 

• Low ‐ computer generated random numbers, table of random numbers, Latin square, coin‐tossing or similar.

• High ‐ even or odd clinic record number or both, clinician judgement, participant preference or similar.

• Unclear ‐ insufficient information about the sequence generation process to permit informed judgement.

Allocation concealment

Was allocation adequately concealed? That is, if the method used to conceal allocation sequence was described sufficiently to tell whether intervention schedules could have been foreseen in advance of, or during the recruitment.   

• Low ‐ drugs of identical appearance in sealed envelopes or containers, central independent department or similar.

• High ‐ rotation, date of birth, non‐opaque envelopes or similar.

• Unclear ‐ randomisation stated but no information given on method used.

Blinding

Was knowledge of the allocated intervention adequately prevented during the trial? That is, the method used to blind participants, personnel, and outcome assessors from knowledge of which intervention a participant received.

• Low ‐ placebo and medication identical.

• High ‐ tablets versus liquid.

• Unclear ‐ blinding stated but no information available on method used.

Incomplete outcome data

Were incomplete data dealt with adequately by researchers?

• Low ‐ no missing outcome data, missing outcome data balanced in numbers across intervention groups, and reasons for dropouts and withdrawals described.

• High ‐ reason for missing outcome data likely to be related to the true outcome.

• Unclear ‐ numbers or reasons for dropouts and withdrawals not described. 

Selective outcome reporting

Are reports of the trial free from suggestion of selective outcome reporting? To assess this, we checked trial protocols, where available, through trial registries and compared the outcomes listed in the protocol with the published report.

• Low ‐ trial protocol is available, published reports include all expected outcomes.

• High ‐ not all of the trial’s prespecified primary outcomes have been reported; one or more reported primary outcomes were not prespecified.

• Unclear ‐ insufficient information to make a judgement.

Other sources of bias

Was the trial apparently free of other problems that could put it at a high risk of bias?

• Low ‐ trial appears to be free of other components that could put it at risk of bias.

• High ‐ there are other factors in the trial that could put it at risk of bias, such as, stopping the trial prematurely, washout inadequacy in cross‐over trials, and extreme baseline imbalance, e.g. inappropriate administration of co‐intervention or use of an insensitive instrument to measure outcomes.

• Unclear ‐ trial may or may not be free of other components that could put it at risk of bias.

Measures of treatment effect

Binary data

We calculated the odds ratio (OR) and risk difference (RD) with 95% confidence intervals (CIs) for binary outcomes. Where there was a positive effect from use of TCA, we calculated the number needed to treat for an additional beneficial outcome (NNTB). We attempted to analyse the following binary outcomes: ADHD, depressive, anxiety, obsessive, and behavioural problem symptom severity expressed as response rates as well as all‐cause treatment discontinuation.

Continuous data

We used both change and endpoint scores. We combined trials using endpoint scores with those using change scores as this method has been shown to be valid (Da Costa 2013). We calculated the standardised mean difference (SMD) and 95% CIs when trials did not use sufficiently similar instruments to measure an outcome; we calculated the mean difference (MD) when trials used the same instrument. We assessed the following continuous outcomes: ADHD, depressive, anxiety, obsessive symptoms, and behavioural problem severity expressed as change of endpoint scores.

Unit of analysis issues

We did not identify any cluster RCTs for inclusion and we did not identify any unit of analysis errors. Had there been cluster‐RCTs that otherwise fit our inclusion criteria, we would not have used them because using this design type for trials of TCAs may have resulted in biased data (Higgins 2011).

To incorporate cross‐over trials, we attempted to use recommendations by Elbourne 2002. However, this was not possible because there was no correlation coefficient between the assessments of the trial periods in the trials with cross‐over design. Therefore, we only used first‐period data from cross‐over trials that fulfilled our inclusion criteria, as proposed in the Cochrane Protocol (Otasowie 2008).

Dealing with missing data

We reported the loss to follow‐up and reasons for missing data where available. Where the reporting of data appeared to be incomplete, we made all reasonable efforts to contact trial authors to request missing data but did not receive any response.

Assessment of heterogeneity

We assessed clinical heterogeneity by comparing the distribution of important participant factors between trials (e.g. age, inclusion of participants who have failed to respond to stimulant medication, patients with comorbid tic disorder). We analysed methodological heterogeneity by reporting variability in the designs of the trials with regards to randomisation concealment, blinding of outcome assessment, losses to follow‐up, treatment type, co‐interventions, and variable drug dose range. Also, we determined the statistical heterogeneity by examining I² (Higgins 2002), a figure which describes the approximate proportion of variation in point estimates that is due to heterogeneity rather than to sampling error. In addition, we employed a Chi² test of homogeneity to determine the strength of evidence that heterogeneity is genuine. The following cut‐offs were used to determine the importance of statistical inconsistency: 0% to 30% ‐ not important, 30% to 60% ‐ moderate, 60% to 90% ‐ substantial, and > 90% ‐ considerable (Higgins 2011).

Assessment of reporting biases

We planned to assess reporting publication bias using funnel plots, but we were unable to do so because insufficient trials were available (see Otasowie 2008 for further information).

Data synthesis

We calculated the SMD for continuous data and we calculated the OR for dichotomous data. We presented all measures with a 95% CI and we pooled data using a random‐effects model throughout.

Subgroup analysis and investigation of heterogeneity

We had planned to conduct subgroup analyses for:

1. Individual TCA (e.g. desipramine, clomipramine, imipramine, amitriptyline).

2. Age groups (younger than 12 years of age, 12 to 18 years of age).

3. Dosage.

We could only perform subgroup analyses for individual TCA because of a lack of variability for age and dosage (see Differences between protocol and review section).

Sensitivity analysis

We conducted the following sensitivity analyses to assess the impact of trial risk of bias according to:

1. Only those with low risk of selection bias (associated with sequence generation or allocation concealment).

2. Only those with low risk of performance bias (associated with issues of blinding).

3. Only those with low risk of attrition bias (associated with completeness of data).

We also performed a post hoc sensitivity analysis after excluding one trial that scored a 'high risk' of reporting bias.

Results

Description of studies

Results of the search

Our initial search of all identified databases returned 1083 citations, 68 of which we deemed worthy of further scrutiny. Of these, only 17 were RCTs of TCAs in the treatment of ADHD in children and adolescents. Following full‐text screening, only six RCTs met our predefined inclusion criteria. We excluded 11 trials because five trials were conducted in hyperactive individuals without a formal diagnosis of ADHD; four assessed the pharmacological effects on cognition and attention but did not assess the primary symptoms of ADHD; one trial was specifically conducted to assess the side effects of TCA in children, and the remaining trial was designed to ascertain the active sites for ADHD medications (see Figure 1).

1.

Study flow diagram

Included studies

Intervention

Six trials reported the effects of three TCAs (desipramine, clomipramine, and nortriptyline) on ADHD symptoms (Garfinkel 1983; Donnelly 1986; Biederman 1989; Singer 1995; Prince 2000; Spencer 2002). In addition, Garfinkel 1983 studied the effect of methylphenidate and Singer 1995 studied clonidine.

The TCA dose used in the interventions varied: desipramine dosage ranged from 3.5 to 4.6 mg/kg/day; clomipramine dosage was 85 mg/day; nortriptyline was 1.8 mg/kg/day; clonidine was 0.19 mg/day; and methylphenidate was 0.75 mg/kg/day. Treatment duration ranged from two to six weeks for desipramine, nine weeks for nortriptyline, and three weeks for clomipramine.

Trial design

Most included trials were of small size, were of heterogeneous design, and assessed multiple drugs such as desipramine, clomipramine, nortriptyline, methylphenidate, and clonidine.

Two trials used a cross‐over design. Garfinkel 1983 compared desipramine, clomipramine, methylphenidate, and placebo, and reported results as means and F‐values, which do not allow for the calculation of effect sizes. Therefore, we could not include this trial in any analysis because we could not combine the results using meta‐analysis due to the way in which they were reported. Singer 1995 compared desipramine, clonidine, and placebo. From this trial, we used the first period of the comparisons desipramine versus placebo and desipramine versus clonidine.

The four remaining trials were of parallel design. Three compared desipramine with placebo (Donnelly 1986; Biederman 1989; Spencer 2002), and one compared nortriptyline with placebo (Prince 2000). Prince 2000 was slightly different from the other trials because instead of investigating the efficacy of nortriptyline, trialists assessed the effect of treatment discontinuation in patients who responded to the medication during a six‐week, open label period.

Setting

All included trials were conducted in outpatient clinics in the USA, except for Donnelly 1986, which was conducted in a day hospital.

Participants

In total, the six trials recruited 216 participants aged between six and 17 years, 90% of whom were boys. The total number of participants in each included trial ranged from 12 to 62, with a mean age of 9.9 years. All participants had normal levels of cognitive functioning based on clinical assessment.

The included trials used three different versions of the DSM criteria; three trials used DSM‐III (Garfinkel 1983; Donnelly 1986; Biederman 1989), one trial used DSM‐III‐R (Singer 1995), and two trials used DSM‐IV (Prince 2000; Spencer 2002).

Two trials used the presence of a comorbid disorder as an inclusion criterion (Singer 1995; Spencer 2002). These conditions were Tourette disorder in Singer 1995 and chronic tic disorder in Spencer 2002. More than half of the participants were previously treated with stimulants.

Outcome measures

The six included trials used a diversity of outcome measures ‐ 30 in total ‐ which made it difficult to make statistical comparisons between treatment regimens. The most commonly used instruments were Conners' ADHD Rating Scales (Garfinkel 1983; Donnelly 1986; Biederman 1989; Spencer 2002), CGI Scale (Biederman 1989; Prince 2000; Spencer 2002), Children's Depression Inventory (Biederman 1989; Prince 2000; Spencer 2002), Subjective Treatment Emergent Symptoms Scale (Donnelly 1986; Biederman 1989), and RCMAS (Prince 2000; Spencer 2002). Tourette Syndrome Severity Scale and Yale Global Tic Severity Scale were both used by trials that investigated the effects of TCA on children with ADHD and comorbid tic disorder.

All trials evaluated the severity of ADHD symptoms but trial authors used different scales and raters. For instance, three trials used a parent‐rated instrument (Biederman 1989; Singer 1995; Spencer 2002) and three trials used a teacher‐rated form (Garfinkel 1983; Donnelly 1986; Biederman 1989). Garfinkel 1983 assessed the effect of the studied interventions on ADHD symptom severity by means of the teacher‐ and child‐care, worker‐rated Conners' Rating Scale. This validated rating scale measures the core symptoms of ADHD, such as hyperactivity, impulsivity, and inattention, as well as three other dimensions, namely sociability, depressive, and self‐esteem.

Three trials assessed the efficacy of TCA (Biederman 1989; Prince 2000; Spencer 2002) using the CGI, which is a clinician‐rated instrument. One trial, Prince 2000, also used CGI as a rating scale.

Two trials included patients with comorbid tic disorder or Tourette disorder, and the trial authors used rating scales specifically validated for assessing these co‐morbidities (Singer 1995; Spencer 2002).

Two trials (Biederman 1989; Spencer 2002) investigated the effect of TCA on depressive symptoms. In addition, Spencer 2002 assessed the effect of TCA on anxiety and Prince 2000 assessed the effect on oppositional defiant disorder (ODD) symptoms. The trialists in Donnelly 1986 assessed teacher‐rated ADHD symptom severity and several neuropsychological functions as well as drug, catecholamine plasma, and urinary levels.

Primary outcomes

ADHD symptoms

Conners' Parent Rating Scale (CPRS): This measures hyperactivity and behavioural problems in children and consists of different indexes. The hyperactivity index is used for this outcome. There are 39 items, which are rated on a four‐point Likert scale ranging from zero (not at all true) to three (very true). Three trials used this scale (Donnelly 1986; Biederman 1989; Singer 1995).

Conners' Teacher Rating Scale (CTRS): This scale is used for the assessment of ADHD, and also measures hyperactivity and behavioural problems in children. It consists of different indexes and the hyperactivity index is used in this outcome. There are 39 items, which are rated on a four‐point Likert scale ranging from zero (not at all true) to three (very true). Garfinkel 1983 and Donnelly 1986 used this scale.

Parent and Teacher CBCL: This assesses a child's competencies and problems using a standardized 113‐item questionnaire. The behaviour profile shows the child's standing on nine behavioural problem scales and three social competence scales; scored on a three‐point scale ranging from zero (not true) to two (often true). This scale was used in Singer 1995.

CGI Scale: This measures the child's overall level of improvement on a seven‐point scale, ranging from one (much worse) to seven (much improved). Three trials used this measure (Biederman 1989; Prince 2000; Spencer 2002).

ADHD Rating Scale: This scale consists of 18 items from the DSM‐IV diagnostic criteria for ADHD. It is rated on a five‐point Likert scale from zero (never) to four (almost always). Prince 2000 and Spencer 2002 used this scale.

Adverse events

Most included trials provided data on adverse events (Donnelly 1986; Biederman 1989; Singer 1995; Prince 2000; Spencer 2002). One trial collected information on adverse events using a checklist (Singer 1995), one trial used open‐ended questions (Spencer 2002), and two trials used the Subjective Treatment Emergent Symptoms Scale (STESS), a clinician‐rated scale that records reported and observed new symptoms (Donnelly 1986; Biederman 1989). Prince 2000 did not report the way adverse events were collected. No serious adverse events were reported in any of the included trials.

Secondary outcomes

General behaviour

CBCL: This assesses a child's competencies and problems using a standardized 113‐item questionnaire. The behaviour profile shows the child's standing on nine behavioural problem scales and three social competence scales; scored on a three‐point scale ranging from zero (not true) to two (often true). Singer 1995 used this scale.

Depression

Children's Depression Inventory: This assesses a child's symptoms of depression using a standardized, 27‐item questionnaire. Three trials used this (Biederman 1989; Prince 2000; Spencer 2002).

Anxiety

RCMAS is a 37‐item scale that is rated with T scores. Prince 2000 and Spencer 2002 used this scale.

Obsessive symptoms

Prince 2000 used CGI to assess obsessive symptoms.

Excluded studies

We excluded 11 studies after full‐paper review. Five were excluded because the participants, who were mostly described as hyperactive, were not formally diagnosed with ADHD (Krakowski 1965; Rapoport 1974; Yepes 1977; Yellin 1978; Werry 1980); four were excluded because the studies did not evaluate the core symptoms of ADHD but mainly assessed the pharmacological effects of TCAs on cognition and attention using non‐disorder‐specific outcome measures such as WISC (Gualtieri 1988; Tellechea 1991; Carlson 1995; Guardiola 1999); one was excluded because it evaluated the side effects of TCA on patients with ADHD (Pataki 1993), and another because it focused on the active sites for ADHD medications (Overtoom 2003).

See the 'Characteristics of excluded studies' section for more details.

Risk of bias in included studies

No trial was free from bias across all items of the Cochrane 'Risk of bias' tool; randomization and allocation concealment were often poorly described.

Allocation

Sequence generation

Sequence generation was adequate in three trials (Garfinkel 1983; Biederman 1989; Spencer 2002) and was unclear in the others (Donnelly 1986; Singer 1995; Prince 2000). Biederman 1989 assigned children randomly by a computer‐generated list. The pharmacy randomised participants using separate, balanced randomisation in Spencer 2002 whilst Garfinkel 1983 used a 4 X 4 Latin square. The remaining trials did not describe the mode of sequence generation.

Allocation concealment

Two trials had adequate allocation concealment (Singer 1995; Spencer 2002), but no further information was provided as regards how allocation concealment was achieved in the remaining four trials (Garfinkel 1983; Donnelly 1986; Biederman 1989; Prince 2000).

Blinding

All six trials reported that participants and their carers, as well as the clinicians involved in medicating the participants, were blinded to allocation to placebo or to intervention phases. However, Prince 2000 did not report how blinding was performed. We scored the risk of performance and detection bias as 'low' in the five trials where medication was reported to have an identical appearance, thus, clinical ratings were obtained under double‐blind conditions (Garfinkel 1983; Donnelly 1986; Biederman 1989; Singer 1995; Spencer 2002). Although Prince 2000 used the term "double‐blind" to describe their trial, the trialists did not specifically report the blinding of the assessors. This is relevant because the participants were assessed using the same tools and rating scales during the open, placebo, and treatment phases; as these measures are largely observation‐based, there is potential for the results to be biased if the assessors were not adequately blinded to which phase each participant was allocated.

Incomplete outcome data

Regarding subjective outcomes, all trials were judged free of attrition bias because discontinuation rate was low or because they consisted of cross‐over studies with a paired analysis, and thus are trial group comparable. All‐cause treatment discontinuation was also rated as having a low risk of attrition bias because this outcome is, by definition, unaffected by this source of bias.

Selective reporting

The majority of the trials were conducted prior to the introduction of trial registration sites that contain protocols and predefined outcome measures; therefore, we assessed whether the outcome measures described in the method of the paper were reported in the results section of the paper.

Most trials were considered free of reporting bias because the published results corresponded to those expected in these types of studies. However, on further scrutiny, we noticed some evidence of selective reporting in two trials. For instance, Prince 2000 did not report the number of patients discontinuing in each trial arm and the trial authors did not offer reasons for patients dropping out. Singer 1995 did not provide all outcome data either. It seems that the trial authors did not report outcomes for all measures of their trials as planned, but only reported significant figures.

Other potential sources of bias

Two trials had a cross‐over design (Garfinkel 1983; Singer 1995); Garfinkel 1983 did not test for the presence of a carry‐over effect whilst Singer 1995 showed that such an effect was likely.

Two trials (Biederman 1989; Prince 2000) were partly sponsored by the manufacturers of desipramine and nortriptyline respectively. Therefore, these trials were deemed to have unclear risk of bias.

Another important source of bias is the duration of treatment. It is important to state that the short duration of treatment, for instance three weeks in Garfinkel 1983, Donnelly 1986, and Prince 2000, is much too brief to adequately assess the benefits and risks of tricyclic pharmacotherapy. This is particularly important when comparing the efficacy of a TCA to a stimulant.

Effects of interventions

See: Table 1; Table 2

We have summarized the effects of desipramine and nortriptyline versus placebo and clonidine in the 'Data and analyses' section. The comparisons of clomipramine versus placebo, and TCA versus methylphenidate could not be analysed using meta‐analytical techniques because the figures reported in the publications did not allow for the calculation of effect sizes. We therefore elected to present a narrative description for these comparisons.

See: Table 1 for the main comparisons of TCA versus placebo for ADHD, and Table 2 for the main comparisons of desipramine versus clonidine.

We present the results of our investigated primary outcome (effect of TCA on core symptoms of ADHD and adverse events) and secondary outcome (change in behavioural problems). We describe our findings across the following domains:

• Changes in the child's ADHD symptom‐related behaviour in the home setting: assessed with CGI (NIMH 1985), CBCL (Achenbach 1991), and CPRS (Conners 1994).

• Changes in the child's ADHD symptom‐related behaviour in the school setting: assessed with CGI (NIMH 1985), CBCL (Achenbach 1991), and CTRS (Conners 1994).

• Changes in the child's general behaviour: assessed with CGI (NIMH 1985) and CBCL (Achenbach 1991).

• Depression: assessed with CDI (Kovacs 1985).

TCA versus placebo

Primary outcomes

All included trials (Garfinkel 1983; Donnelly 1986; Biederman 1989; Singer 1995; Prince 2000; Spencer 2002) investigated the effect of a TCA on ADHD symptoms. Nevertheless, we could only pool data from more than one TCA for one trial outcome, namely the proportion of responders using the CGI. For this outcome we found TCA to be more efficacious than placebo in three trials (Biederman 1989; Prince 2000; Spencer 2002) (OR 18.50, 95% CI 6.29 to 54.39, 125 participants; Analysis 1.1). The calculated NNTB was two, from 1.33 to 2.04. This finding was consistent for both TCAs (desipramine and nortriptyline). There was no evidence of statistical heterogeneity.

1.1. Analysis.

Comparison 1 TCA versus placebo, Outcome 1 CGI (response rate).

The efficacy of desipramine on ADHD symptoms as assessed by parents was available for three trials (Biederman 1989; Singer 1995; Spencer 2002) but one trial (Singer 1995) used a visual analogue scale (VAS) instead of a rating scale, which made it difficult to pool the results of all included trials. So, we have reported the results on the efficacy of TCA on ADHD symptoms separately for trials using rating scales (Analysis 1.2; Analysis 1.3; Analysis 1.4) and those using VAS (Analysis 1.5).

1.2. Analysis.

Comparison 1 TCA versus placebo, Outcome 2 Efficacy for ADHD symptoms (parents using a rating scale).

1.3. Analysis.

Comparison 1 TCA versus placebo, Outcome 3 Efficacy for ADHD symptoms (parents using a rating scale) (response rate).

1.4. Analysis.

Comparison 1 TCA versus placebo, Outcome 4 Efficacy for ADHD symptoms (teachers using a rating scale).

1.5. Analysis.

Comparison 1 TCA versus placebo, Outcome 5 Efficacy for ADHD symptoms (parents using VAS).

In two trials (Biederman 1989; Spencer 2002), desipramine was more efficacious than placebo for reducing ADHD symptoms as assessed by clinicians (OR 26.41, 95% CI 7.41 to 94.18, I² = 0%, 103 participants; Analysis 1.1.1, part of Analysis 1.1) and parents (SMD ‐1.42, 95% CI ‐1.99 to ‐0.85, I² = 37%, 99 participants; Analysis 1.2). Desipramine was also more efficacious than placebo for reducing ADHD symptoms as assessed by teachers (SMD ‐0.97, 95% CI ‐1.66 to ‐0.28, I² = 53%, 2 trials (Biederman 1989; Donnelly 1986), 89 participants; Analysis 1.4). (The calculated NNTB to achieve a CGI ‐ improvement score of one or two was two (95% CI 1.3 to 2.0; I² = 0%). Desipramine also outperformed placebo when CGI was expressed as a continuous variable (Analysis 1.6). Furthermore, desipramine improved child general behaviour as assessed by teachers using the CBCL scale (Analysis 1.7; Analysis 1.8; Analysis 1.9; Analysis 1.10; Analysis 1.11).

1.6. Analysis.

Comparison 1 TCA versus placebo, Outcome 6 CGI (% of change from baseline).

1.7. Analysis.

Comparison 1 TCA versus placebo, Outcome 7 Child general behaviour (CBCL ‐ hyperactive subscale).

1.8. Analysis.

Comparison 1 TCA versus placebo, Outcome 8 Child general behaviour (CBCL ‐ nervous / overactive subscale).

1.9. Analysis.

Comparison 1 TCA versus placebo, Outcome 9 Child general behaviour (CBCL ‐ unpopular subscale).

1.10. Analysis.

Comparison 1 TCA versus placebo, Outcome 10 Child general behaviour (CBCL ‐ anxious subscale).

1.11. Analysis.

Comparison 1 TCA versus placebo, Outcome 11 Child general behaviour (CBCL ‐ obsessive compulsive subscale).

Donnelly 1986 observed rapid onset of clinical response in the desipramine group; this is similar to the stimulant effect and might suggest that the effect is not mediated through the down regulation of receptors. However, this type of response was not replicated in Biederman 1989 where the onset of significant clinical improvement occurred between three and four weeks after receiving the TCA.

One trial investigated the efficacy of nortriptyline on ADHD symptoms (Prince 2000). This trial found that patients continuing with nortriptyline showed lower scores of investigator‐rated ADHD symptoms and ADHD‐CGI (Analysis 1.1.2, part of Analysis 1.1).

Only one trial, Garfinkel 1983, investigated clomipramine efficacy. Clomipramine was found superior to placebo in reducing ADHD symptoms using the CTRS.

Regarding safety, no serious adverse events were reported for any TCA studied. Patients treated with desipramine had significantly higher rates of appetite suppression compared with placebo. There were mild but statistically significant increases in diastolic blood pressure and pulse rates in desipramine‐treated participants. Other reported adverse effects included dry mouth, abdominal discomfort, sedation, tiredness, headaches, drowsiness, confusion, blurred vision, diaphoresis, constipation, and urinary retention. The most common adverse events of clomipramine were dry mouth, constipation, and headache. No clinically significant electrocardiogram (ECG) changes were reported. Little information on clomipramine safety was available and it was only reported to have raised blood pressure and heart rate.

Secondary outcomes

We assessed the effect of TCA on depressive symptoms in the trials that investigated desipramine and clomipramine. Two trials (Biederman 1989; Spencer 2002) investigated the effect of desipramine on depressive symptoms, but only Biederman 1989 reported the results in a way suitable for meta‐analysis (Analysis 1.12.1, part of Analysis 1.12). This trial found that desipramine improved depressive symptoms compared with placebo. However, this finding was not replicated by Spencer 2002; the trial authors found no statistically significant differences in this outcome. One trial (Garfinkel 1983) investigated the efficacy of clomipramine on depressive symptoms. In this trial, clomipramine was significantly superior to placebo in ameliorating depressive symptoms.

1.12. Analysis.

Comparison 1 TCA versus placebo, Outcome 12 Depressive symptoms.

The effect of TCA on anxiety symptoms was only studied for desipramine in two clinical trials (Singer 1995; Spencer 2002). While Singer 1995 found that desipramine improved anxiety symptoms assessed with the CBCL‐anxiety subscale which did not reach statistical significance (Analysis 1.10), Spencer 2002 found no improvement compared with placebo.

Three trials investigated the effects of desipramine (Singer 1995; Spencer 2002) and nortriptyline (Prince 2000) on ODD symptoms. Desipramine was not more efficacious than placebo for reducing ODD symptoms. Conversely, a positive effect was found for nortriptyline. Participants who responded to nortriptyline treatment maintained their clinical improvement in ODD symptoms, whilst those randomised to the placebo group had a significant deterioration in these symptoms (Analysis 1.13).

1.13. Analysis.

Comparison 1 TCA versus placebo, Outcome 13 CGI ‐ ODD symptoms.

Three trials (Donnelly 1986; Biederman 1989; Spencer 2002) investigating the efficacy of desipramine analysed the effect of this TCA on all‐cause treatment discontinuation. There were no differences between desipramine and placebo on this outcome (Analysis 1.14.1, part of Analysis 1.14), although statistical heterogeneity was statistically significant (P = 0.05). Treatment discontinuation was not reported in the trials that investigated clomipramine and nortriptyline.

1.14. Analysis.

Comparison 1 TCA versus placebo, Outcome 14 All‐cause treatment discontinuation.

TCA versus active medications

One trial, Singer 1995, compared desipramine with clonidine in children and adolescents with ADHD and comorbid Tourette syndrome. This trial found desipramine to be more efficacious than clonidine in reducing ADHD symptoms (Analysis 2.1; Table 2).

2.1. Analysis.

Comparison 2 Desipramine versus clonidine, Outcome 1 Efficacy for ADHD symptoms (parent‐rated symptoms using VAS).

We could only analyse data on children's global behaviour using the CBCL, which were only reported for boys of various age ranges. Desipramine was superior to clonidine in the hyperactive subscale of the maternal‐completed CBCL for boys aged six to 11 years (Analysis 2.2), and in the nervous or overactive (Analysis 2.3), and obsessive compulsive subscales of the teacher‐completed CBCL (Analysis 2.6) for boys older than 12 years of age. There were no differences between the unpopular (Analysis 2.4) and anxious subscales (Analysis 2.5).

2.2. Analysis.

Comparison 2 Desipramine versus clonidine, Outcome 2 Child general behaviour (CBCL ‐ hyperactive subscale).

2.3. Analysis.

Comparison 2 Desipramine versus clonidine, Outcome 3 Child general behaviour (CBCL ‐ nervous / overactive subscale).

2.6. Analysis.

Comparison 2 Desipramine versus clonidine, Outcome 6 Child general behaviour (CBCL ‐ obsessive compulsive subscale).

2.4. Analysis.

Comparison 2 Desipramine versus clonidine, Outcome 4 Child general behaviour (CBCL ‐ unpopular subscale).

2.5. Analysis.

Comparison 2 Desipramine versus clonidine, Outcome 5 Child general behaviour (CBCL ‐ anxious subscale).

Regarding safety, no differences were found between desipramine and clonidine as regards the number of patients experiencing side effects. The trial authors reported that almost half of the children who completed the entire protocol reported having at least one "drug‐related" problem in the placebo phase. Although the difference between drug and placebo phases was statistically significant, no single side effect correlated significantly with a specific drug.

Another trial, Garfinkel 1983, compared desipramine against clomipramine and methylphenidate. Whilst methylphenidate was found to be more efficacious in reducing ADHD symptom severity than TCA, no differences were found between clomipramine and desipramine, and placebo. Conversely, clomipramine was significantly superior to desipramine and methylphenidate in ameliorating depressive symptoms.

Sensitivity analyses

We performed three sensitivity analyses. We limited the first sensitivity analysis (Analysis 3.1; Analysis 3.2; Analysis 3.3; Analysis 3.4; Analysis 3.5) to the one trial that scored 'low risk' on selection bias (Spencer 2002). The results were consistent with the primary analyses; desipramine reduced ADHD symptom severity but did not reduce treatment discontinuation. In the second (Analysis 4.1 to Analysis 4.13) and third (Analysis 5.1 to Analysis 5.13) sensitivity analyses, we removed the trial that did not score 'low risk' on the domains 'performance bias' and 'detection bias' (Prince 2000). Removing this trial (TCA nortriptyline) did not result in any change to the results obtained in the primary analysis of desipramine.

3.1. Analysis.

Comparison 3 TCA versus placebo: sensitivity analysis 1, Outcome 1 Efficacy for ADHD symptoms (parents using a rating scale).

3.2. Analysis.

Comparison 3 TCA versus placebo: sensitivity analysis 1, Outcome 2 Efficacy for ADHD symptoms (parents using a rating scale) (response rate).

3.3. Analysis.

Comparison 3 TCA versus placebo: sensitivity analysis 1, Outcome 3 CGI.

3.4. Analysis.

Comparison 3 TCA versus placebo: sensitivity analysis 1, Outcome 4 CGI (response rate).

3.5. Analysis.

Comparison 3 TCA versus placebo: sensitivity analysis 1, Outcome 5 All‐cause treatment discontinuation.

4.1. Analysis.

Comparison 4 TCA versus placebo: sensitivity analysis 2, Outcome 1 Efficacy for ADHD symptoms (parents using a rating scale).

4.13. Analysis.

Comparison 4 TCA versus placebo: sensitivity analysis 2, Outcome 13 All‐cause treatment discontinuation.

5.1. Analysis.

Comparison 5 TCA versus placebo: sensitivity analysis 3, Outcome 1 Efficacy for ADHD symptoms (parents using a rating scale).

5.13. Analysis.

Comparison 5 TCA versus placebo: sensitivity analysis 3, Outcome 13 All‐cause treatment discontinuation.

Finally, we performed one unplanned sensitivity analysis (Analysis 6.1 to Analysis 6.8) after removing one trial that was at 'high risk' of 'reporting bias' (Singer 1995). Again, the main results did not change after running this analysis. Desipramine demonstrated consistency in its ability to effectively reduce ADHD symptom severity.

6.1. Analysis.

Comparison 6 TCA versus placebo: sensitivity analysis 4, Outcome 1 Efficacy for ADHD symptoms (parents using a rating scale).

6.8. Analysis.

Comparison 6 TCA versus placebo: sensitivity analysis 4, Outcome 8 All‐cause treatment discontinuation.

Discussion

Summary of main results

We only identified six small, short‐term RCTs for inclusion in this Cochrane Review. These trials evaluated the effects of three TCAs (desipramine, clomipramine, and nortriptyline) on core symptoms of ADHD in children and adolescents. Most of the participants had previously used stimulant medication, and two of the six trials listed coexistence of tic or Tourette disorder as an inclusion criterion.

This review shows that TCA, particularly desipramine, improved the core symptoms of ADHD as assessed by parents, clinicians, and teachers; the effect size was moderate‐to‐large. In addition, desipramine and nortriptyline were well‐tolerated by trial participants. Most adverse events were mild and no serious adverse events were reported. All‐cause treatment discontinuation was similar for desipramine and placebo. No information was reported as regards clomipramine's safety and treatment discontinuation.

TCAs were also compared to active medications such as clonidine and methylphenidate. One trial compared desipramine to clonidine in children and adolescents with ADHD and comorbid Tourette disorder. Desipramine was both more efficacious for reducing hyperactivity symptoms than clonidine and had comparable tolerability. It improved ADHD symptoms in patients with co‐occurring tics and Tourette disorder. However, this finding should be viewed with caution because of its limited clinical applicability. For instance, in clinical practice it may take several weeks or months for clonidine to become effective and a six‐week trial of this medication would be insufficient to evaluate its efficacy (Jain 2011). Conversely, another trial (Garfinkel 1983) found methylphenidate to be more efficacious for reducing ADHD symptoms than desipramine and clomipramine.

Overall completeness and applicability of evidence

Relatively few RCTs have investigated the effectiveness of TCAs in children and adolescents with ADHD, and those which have, used different outcome measures making it difficult to compare results. For instance, instruments used to determine the efficacy of TCA for ADHD varied considerably across trials hindering their combination in a meta‐analysis. Therefore, the major limitation was the small number of trials, which were of small sample sizes. Small sample sizes increase the likelihood of type II error, that is, there is insufficient power to detect a significant change where one exists. Type II error is usually resolved by meta‐analysis, which is only possible when a sufficient number of studies use the same outcome measures. Given that scarcity of data results in imprecise pooled effect estimates which, in some occasions, range from small to large efficacy, it is possible that the results of this systematic review could change substantially with the publication of a single, larger, and more robust new trial showing different results.

There was also a predominance of pre‐adolescent Caucasian boys, and the majority of the participants were from urban regions in the United States, thus limiting the generalisability of these findings. Consequently, further studies are needed to confirm these findings in other races and settings. There were no TCA studies of patients diagnosed with ADHD using DSM‐V (APA 2013) criteria as this systematic review predated its publication.  

ADHD commonly presents with comorbid conditions such as mood disorders, conduct problems, substance misuse, and neurodevelopmental disorders, including tic disorder or Tourette syndrome (APA 2000). The presence of comorbid conditions in a patient with ADHD often makes treatment more challenging as the choice of medication is usually influenced by the nature of the patient's coexisting condition, and the degree of functional impairment caused by these conditions. Two trials included in this review listed comorbid tic or Tourette disorder as inclusion criteria. This is particularly relevant because children with ADHD have an increased risk of development of a comorbid tic disorder. Studies have shown that almost a third of children with ADHD present with coexisting tics or Tourette disorder (Spencer 1999; Pringsheim 2011) so any medication prescribed should be effective in ameliorating these symptoms if one is to minimise the use of multiple medications.

Treatment with TCA was suitably tolerated as only mild adverse events were reported. It is notable that all included trials showed evidence of increased heart rate in most of the participants who were administered a TCA. This adverse event is important because there have been at least four reported cases of unexplained sudden deaths in children receiving desipramine for treatment of ADHD, but it remains uncertain whether there is a causal link between desipramine and sudden death.

The duration of each included trial was relatively short, ranging from two to six weeks and there was no medium‐term or long‐term follow‐up in any of the trials. Given that ADHD is a chronic condition where treatment is likely to be continued for a long period of time, we cannot guarantee that the efficacy and safety reported in this Cochrane Review will be comparable in the long‐term.

There are many types of TCAs used in the treatment of psychiatric conditions but this systematic review, based on predefined criteria, investigated only three TCAs (desipramine, clomipramine, and nortriptyline) with desipramine being the most frequently studied. Therefore, the findings presented in this review should be evaluated in the light of their methodological limitations and should not be extrapolated to other medications of the same pharmacological group.

Finally, it is important to emphasize that ADHD is a heterogeneous condition with considerable comorbidity so pharmacological intervention must be seen within the context of a holistic treatment plan. The decision to use medication must be carefully considered and tailored to the needs of the individual.

Quality of the evidence

The overall risk of bias of each trial was judged to be low in one trial whilst another trial was scored as having 'high risk' of bias because the risk of selective reporting was considered to be high (see Figure 2; Figure 3). We performed four sensitivity analyses and the results did not change significantly from the primary one. These findings provide strength to the results of this systematic review.

2.

Risk of bias summary: review authors' judgements about each risk of bias item for each included trial.

3.

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included trials.

Overall, the quality of the evidence ranged from very low to low fundamentally because few small trials have studied the efficacy and safety of TCA for ADHD. Therefore, the statistical power was also low and the effects calculated could change significantly with the publication of one single, clinical trial. Furthermore the statistical heterogeneity was substantial in one outcome (all cause treatment discontinuation). Thus, the confidence with our findings for this particular outcome is very low.

Potential biases in the review process

Given that only six trials met our inclusion criteria, we could not investigate the possibility of publication bias using funnel plots; more trials are needed to make this a viable statistical tool. Therefore, we could not rule out the possibility that the publication of studies with positive results could bias the findings of this Cochrane Review. Nevertheless, TCAs have little economic interest so it is unlikely that negative studies have been hidden.

Agreements and disagreements with other studies or reviews

This Cochrane Review shows evidence of the efficacy of TCA, particularly desipramine, clomipramine, and nortriptyline in the treatment of ADHD in children and adolescents. The results of our review are similar to that by Jadad 1999, which found desipramine to be more effective than placebo in the treatment of ADHD in children and adolescents. A recent Cochrane Review of pharmacological treatment for ADHD in children with comorbid tic disorder also demonstrated the effectiveness of the medication on core symptoms of ADHD (Pringsheim 2011).

Findings from the National Institute of Mental Health (NIMH) research on the treatment of ADHD: the Multimodal Treatment Study (MTA) (MTA 1999) and the Multimodal Psychosocial Treatment Study (MPT) (MTA 2004) indicated that the comprehensive and standardised medication management group showed a relatively superior outcome compared with intensive behavioural treatment alone or routine community treatment (Klein 2004). In a systematic review of the efficacy of different management strategies in ADHD, Klassen 1999 concluded that medication‐only therapy seemed most efficacious although they acknowledged that there might have been a bias in the way the ADHD trial cohorts were assembled, treated, and assessed.

Authors' conclusions

Implications for practice.

TCAs, particularly desipramine, appear efficacious for treating ADHD in children and adolescents in the short term. However, it is relevant to mention that their use in clinical practice might be limited by the infrequency with which they have been studied.

Whilst the focus of this Cochrane Review is on the effect of TCA in the treatment of ADHD in children and adolescents, it is important to point out that there are newer and safer alternative medications available that have been studied more comprehensively than TCA. First‐line pharmacological interventions, including stimulants (methylphenidate and amphetamine derivatives) and non‐stimulants (atomoxetine and alpha agonists), have been studied extensively. Given the availability of these non‐stimulants, it seems likely that TCAs may be considered as third‐line medication for ADHD treatment.

Implications for research.

The large numbers of different outcome measures used in the included trials made it difficult to make a meaningful comparison of data across the trials. All trials focused more on efficacy (i.e. how well TCA works in clinical trials) than on effectiveness (i.e. how well TCA works in 'real world' clinical practice). We hope that future research will assess the effectiveness of TCA in the long‐term, as well as the impact on the quality of life of the children on these medications. Such research should involve large sample sizes, diverse ethnicity, and use standardised outcome measures to facilitate meta‐analyses.

Given the paucity of high quality trials included in this Cochrane Review, we would suggest that future clinical trials include alternative study designs such as longitudinal multicentre observational studies. These types of studies have the potential to yield more generalisable results pertinent to clinic populations and might be more appropriate for evaluating the effects of TCAs on ADHD than placebo‐ or comparator‐controlled trials.

Desipramine has been found to be ineffective in the treatment of isolated depression in children and adolescents (Hazell 2013), but it was found to be effective when used in young people with ADHD and comorbid depression (Biederman 1989). This might suggest that TCAs have a different mechanism of action or preferential effects on receptor subtypes when ADHD coexists with depression. A more detailed investigation is needed to understand the pathophysiology of depression in patients with ADHD and the mechanism of action of TCA in this group of patients.

History

Protocol first published: Issue 1, 2008
 Review first published: Issue 9, 2014

Date	Event	Description
15 September 2008	Amended	We converted to the new Cochrane Review format.

Appendices

Appendix 1. Search strategies

Cochrane Central Register of ControlledTrials (CENTRAL), part of the Cochrane Library

2012 (Issue 6 of 12) searched 28 June 2012, all available years [58 records].
 2013 (Issue 9 of 12 ) searched 26 September 2013, limited to PY 2012 to 2013 [no records].

#1MeSH descriptor Attention Deficit Disorder with Hyperactivity explode all trees
 #2(ADHD OR ADDH OR ADHS OR AD NEXT HD OR HKD)
 #3(attention* or behav*) NEAR/3 (defic* or dysfunc* or disorder*)
 #4(disrupt* near/3 disorder*) or (disrupt* near/3 behav*) or (defian* near/3 disorder*) or (defian* near/3 behav*)
 #5impulsiv* or inattentiv* or inattention*
 #6hyperkin* OR hyper NEXT kin*
 #7hyperactiv* or hyper NEXT activ*
 #8(minimal near/3 brain near/3 (disorder* or dysfunct* or damage*))
 #9(#1 OR #2 OR #3 OR #4 OR #5 OR #6 OR #7 OR #8)
 #10MeSH descriptor Adolescent explode all trees
 #11child near "MESH check words"
 #12child* or boy* or girl* or teen* or adolescen* or young NEXT people
 #13(#10 OR #11 OR #12)
 #14MeSH descriptor Antidepressive Agents, Tricyclic explode all trees
 #15tca
 #16amitriptyline
 #17MeSH descriptor Amoxapine explode all trees
 #18amoxapine
 #19clomipramine
 #20desipramine or dosulepin or dothiepin or doxepin
 #21imipramine or Iprindole
 #22lofepramine
 #23opipramol OR protriptyline
 #24nortriptyline
 #25trimipramine
 #26tricyclic*
 #27(#14 OR #15 OR #16 OR #17 OR #18 OR #19 OR #20 OR #21 OR #22 OR #23 OR #24 OR #25 OR #26)
 #28(#9 AND #13 AND #27)
 
 Ovid MEDLINE(R)

1946 to June Week 3 2012 searched 28 June 2012 [316 records].
 1946 to September Week 3 2013 searched 26 September 2013, limited to ED=20120601‐2013092631 [6 records].

1 "attention deficit and disruptive behavior disorders"/ or attention deficit disorder with hyperactivity/ or conduct disorder/
 2 ADHD.tw.
 3 ADDH.tw.
 4 ADHS.tw.
 5 ("AD/HD" or hkd).tw.
 6 ((attention$ or behav$) adj3 (defic$ or dysfunc$ or disorder$)).tw.
 7 ((disrupt$ adj3 disorder$) or (disrupt$ adj3 behav$) or (defian$ adj3 disorder$) or (defian$ adj3 behav$)).tw.
 8 (impulsiv$ or inattentiv$ or inattention$).tw.
 9 hyperkinesis/
 10 (hyperkin$ or hyper‐kin$).tw.
 11 (minimal adj3 brain adj3 (disorder$ or dysfunct$ or damage$)).tw.
 12 (hyperactiv$ or hyper‐activ$).tw.
 13 or/1‐12
 14 exp Antidepressive Agents, Tricyclic/
 15 amoxapine/
 16 tca.tw.
 17 amitriptyline.tw.
 18 amoxapine.tw.
 19 (clomipramine or c?lorimipramine).tw.
 20 desipramine.tw.
 21 dosulepin.tw.
 22 dothiepin.tw.
 23 doxepin.tw.
 24 imipramine.tw.
 25 iprindole.tw.
 26 lofepramine.tw.
 27 maprotiline.tw.
 28 nortriptyline.tw.
 29 Opipramol.tw.
 30 Protriptyline.tw.
 31 trimipramine.tw.
 32 tricyclic$.tw.
 33 or/14‐32
 34 child/
 35 adolescent/
 36 (child$ or boy$ or girl$ or teen$ or adolescen$ or young people).tw.
 37 or/34‐36
 38 randomized controlled trial.pt.
 39 controlled clinical trial.pt.
 40 randomi#ed.ab.
 41 placebo$.ab.
 42 drug therapy.fs.
 43 randomly.ab.
 44 trial.ab.
 45 groups.ab.
 46 or/38‐45
 47 exp animals/ not humans.sh.
 48 46 not 47
 49 13 and 33 and 37 and 48
 
 Ovid MEDLINE(R) In‐Process & Other Non‐Indexed Citations

June 27, 2012 searched 28 June 2012 [9 records].
 September 25, 2013 searched 26 September 2013 [10 records].

1 ADHD.tw.
 2 ADDH.tw.
 3 ADHS.tw.
 4 ("AD/HD" or hkd).tw.
 5 ((attention$ or behav$) adj3 (defic$ or dysfunc$ or disorder$)).tw.
 6 ((disrupt$ adj3 disorder$) or (disrupt$ adj3 behav$) or (defian$ adj3 disorder$) or (defian$ adj3 behav$)).tw.
 7 (impulsiv$ or inattentiv$ or inattention$).tw.
 8 (hyperkin$ or hyper‐kin$).tw.
 9 (minimal adj3 brain adj3 (disorder$ or dysfunct$ or damage$)).tw.
 10 (hyperactiv$ or hyper‐activ$).tw.
 11 tca.tw.
 12 amitriptyline.tw.
 13 amoxapine.tw.
 14 (clomipramine or c?lorimipramine).tw.
 15 desipramine.tw.
 16 dosulepin.tw.
 17 dothiepin.tw.
 18 doxepin.tw.
 19 imipramine.tw.
 20 Iprindole.tw.
 21 lofepramine.tw.
 22 maprotiline.tw.
 23 nortriptyline.tw.
 24 Opipramol.tw.
 25 Protriptyline.tw.
 26 trimipramine.tw.
 27 tricyclic$.tw.
 28 (child$ or boy$ or girl$ or teen$ or adolescen$ or young people).tw.
 29 or/1‐10
 30 or/11‐27
 31 28 and 29 and 30
 
 Embase (Ovid)

1980 to 2012 Week 25 searched 28 June 2012 [609 records].
 1980 to 2013 Week 38, limited to EM=201226‐201338, searched 26 September 2013 [32 records].

1 attention deficit disorder/
 2 hyperactivity/
 3 conduct disorder/
 4 ADHD.tw.
 5 ADDH.tw.
 6 ADHS.tw.
 7 "AD/HD".tw.
 8 ((attention$ or behav$) adj3 (defic$ or dysfunc$ or disorder$)).tw.
 9 ((disrupt$ adj3 disorder$) or (disrupt$ adj3 behav$) or (defian$ adj3 disorder$) or (defian$ adj3 behav$)).tw.
 10 (impulsiv$ or inattentiv$ or inattention$).tw.
 11 hyperkine$.tw.
 12 (minimal adj3 brain adj3 (disorder$ or dysfunct$ or damage$)).tw.
 13 hyperactiv$.tw.
 14 or/1‐13
 15 exp Tricyclic Antidepressant Agent/
 16 tca.tw.
 17 amitriptyline.tw
 18 amoxapine.tw.
 19 (clomipramine or c?lorimipramine).tw.
 20 desipramine.tw.
 21 dosulepin.tw.
 22 dothiepin.tw.
 23 doxepin.tw.
 24 imipramine.tw.
 25 iprindole.tw.
 26 lofepramine.tw.
 27 maprotiline.tw.
 28 nortriptyline.tw.
 29 Opipramol.tw.
 30 Protriptyline.tw.
 31 trimipramine.tw.
 32 tricyclic$.tw.
 33 or/15‐32
 34 child/
 35 adolescent/
 36 (child$ or boy$ or girl$ or teen$ or adolescen$ or young people).tw.
 37 or/34‐36
 38 exp Clinical trial/
 39 Randomized controlled trial/
 40 Randomization/
 41 Single blind procedure/
 42 Double blind procedure/
 43 Crossover procedure/
 44 Placebo/
 45 Randomi#ed.tw.
 46 RCT.tw. (9412)
 47 (random$ adj3 (allocat$ or assign$)).tw.
 48 randomly.ab.
 49 groups.ab.
 50 trial.ab.
 51 ((singl$ or doubl$ or trebl$ or tripl$) adj3 (blind$ or mask$)).tw.
 52 Placebo$.tw.
 53 Prospective study/
 54 (crossover or cross‐over).tw.
 55 prospective.tw.
 56 or/38‐55
 57 14 and 33 and 37 and 56
 
 CINAHL Plus (EBSCOhost)

1937 to current, all available years, searched 28 June 2012 [26 records].
 1937 to current, limited to EM=20120601 onwards, searched 26 September 2013 [2 records].

S50 S34 and S49
 S49 S35 or S36 or S37 or S38 or S39 or S40 or S41 or S42 or S43 or S44 or S45 or S46 or S47 or S48
 S48 TI (evaluat* study or evaluat* research) or AB (evaluate* study or evaluat* research) or TI (effectiv* study or effectiv* research) or AB (effectiv* study or effectiv* research) OR TI (prospectiv* study or prospectiv* research) or AB(prospectiv* study or prospectiv* research) or TI (follow‐up study or follow‐up research) or AB (follow‐up study or follow‐up research)
 S47 placebo*
 S46 crossover* or "cross over*"
 S45 (MH "Crossover Design") .
 S44 (tripl* N3 mask*) or (tripl* N3 blind*)
 S43 (trebl* N3 mask*) or (trebl* N3 blind*)
 S42 (doubl* N3 mask*) or (doubl* N3 blind*)
 S41 (singl* N3 mask*) or (singl* N3 blind*)
 S40 (clinic* N3 trial*) or (control* N3 trial*)
 S39 (random* N3 allocat* ) or (random* N3 assign*)
 S38 randomis* or randomiz*
 S37 (MH "Meta Analysis")
 S36 (MH "Clinical Trials+")
 S35 MH random assignment
 S34 S30 and S33
 S33 S31 or S32
 S32 child* or boy* or girl* or adolescen* or teen* or young people
 S31 MH "CHILD" OR MH "adolescence"
 S30 S12 and S29
 S29 S13 or S14 or S15 or S16 or S17 or S18 or S19 or S20 or S21 or S22 or S23 or S24 or S25 or S26 or S27 or S28
 S28 trimipramine
 S27 Opipramol
 S26 Nortriptyline
 S25 lofepramine or maprotiline
 S24 Iprindole
 S23 Imipramine
 S22 doxepin or dothiepin
 S21 desipramine
 S20 dosulepin
 S19 clomipramine or clorimipramine or chlorimipramine
 S18 amoxapine
 S17 amitriptyline
 S16 Protriptyline
 S15 TCA
 S13 MH "Antidepressive Agents, Tricyclic+"
 S12 S1 or S2 or S3 or S4 or S5 or S6 or S7 or S8 or S9 or S10 or S11
 S11 disruptiv*
 S10 (impulsiv* or inattentiv* or in‐attentiv* or inattention* or in‐attention*)
 S9 (behav* N3 (dysfunc* or disorder*))
 S8 (attention* N3 (deficit* or disorder* or dysfunc*))
 S7 (minimal N3 brain* N3 (damag* or disorder* or dysfunc*))
 S6 hyperkin* or hyper‐kin*
 S5 (MH "Hyperkinesis")
 S4 hyperactiv* or hyper‐activ* or overactiv* or over‐activ*
 S3 adhd or addh or adhd or "ad/hd" or hkd
 S2 (MH "Child Behavior Disorders")
 S1 (MH "Attention Deficit Hyperactivity Disorder")
 
 PsycINFO (OVID)

1806 to June Week 3 2012 searched 28 June 2012 [111 records].
 1806 to September Week 4 2013 searched 26 September 2013 [4 records].

1 exp attention deficit disorder/
 2 exp Behavior Problems/
 3 Impulsiveness/
 4 hyperkinesis/
 5 adhd.tw.
 6 addh.tw.
 7 adhs.tw.
 8 "ad/hd".tw.
 9 hyperactiv$.tw.
 10 hyper‐activ$.tw.
 11 overactiv$.tw.
 12 over‐activ$.tw.
 13 hyperkin$.tw.
 14 hyper‐kin$.tw.
 15 hkd.tw.
 16 (minimal adj3 brain$ adj3 (damag$ or disorder$ or dysfunc$)).tw.
 17 (attention$ adj3 (deficit$ or disorder$ or dysfunc$)).tw.
 18 (behav$ adj3 (dysfunc$ or disorder$)).tw.
 19 disruptiv$.tw.
 20 (impulsiv$ or inattentiv$ or in‐attentiv$ or in‐attention$ or inattention$).tw.
 21 or/1‐20
 22 exp tricyclic antidepressant drugs/
 23 TCA.tw.
 24 amoxapine.tw.
 25 amitriptyline.tw.
 26 (clomipramine or c?lorimipramine).tw.
 27 desipramine.tw.
 28 dosulepin.tw.
 29 dothiepin.tw.
 30 doxepin.tw.
 31 imipramine.tw.
 32 iprindole.tw.
 33 lofepramine.tw.
 34 maprotiline.tw.
 35 nortriptyline.tw.
 36 opipramol.tw.
 37 protriptyline.tw.
 38 trimipramine.tw.
 39 tricyclic$.tw.
 40 or/22‐39
 41 (adolescence 13 17 yrs or childhood birth 12 yrs or school age 6 12 yrs or young adulthood 18 29 yrs).ag.
 42 (child$ or boy$ or girl$ or teen$ or adolescen$ or young people).tw.
 43 41 or 42
 44 clinical trials/
 45 (randomis* or randomiz*).tw.
 46 (random$ adj3 (allocat$ or assign$)).tw.
 47 ((clinic$ or control$) adj trial$).tw.
 48 ((singl$ or doubl$ or trebl$ or tripl$) adj3 (blind$ or mask$)).tw.
 49 (crossover$ or "cross over$").tw.
 50 random sampling/
 51 Experiment Controls/
 52 Placebo/
 53 placebo$.tw.
 54 exp program evaluation/
 55 treatment effectiveness evaluation/
 56 ((effectiveness or evaluat$) adj3 (stud$ or research$)).tw.
 57 or/44‐56
 58 21 and 40 and 57
 
 SCI and SSCI (Web of Science)

SCI 1970 to 27 June 2012 searched 28 June 2012 [317 records]; SSCI 1970 to 27 June 2012 searched 28 June 2012 [309 records].
 SCI 1 June 2012 to 25 September 2013 searched 26 September 2013 [8 records]; SSCI 1 June 2012 to 25 September 2013, searched 26 September 2013 [4 records].

CPCI‐S 1990 to June 27 2012 searched 28 June 2012 [25 records]; CPCI‐SS&H 1990 to June 27 2012 searched 28 June 2012 [16 records].
 CPCI‐S 1 June 2012 to 25 September 2013 searched 26 September 2013 [no records]; CPCI‐SS&H 1 June 2012 to 25 September 2013 searched 26 September 2013 [no records].

#9 #8 AND #7
 DocType=All document types; Language=All languages;
 #8 TS=(child* or boy* or girl* or teen* or adolescen* or "young people " or youth*)
 DocType=All document types; Language=All languages;
 #7 #6 AND #5
 DocType=All document types; Language=All languages;
 #6 TS= (tca OR amitriptyline OR amoxapine OR clomipramine or clorimipramine OR chlorimipramine or desipramine OR dosulepin OR dothiepin OR doxepin OR imipramine OR iprindole OR lofepramine OR maprotiline OR nortriptyline OR Opipramol OR Protriptyline OR trimipramine OR tricyclic*)
 DocType=All document types; Language=All languages;
 #5 #4 OR #3 OR #2 OR #1
 DocType=All document types; Language=All languages;
 #4 TS= ( impulsiv* or inattentiv* or inattention* or hyperactiv* or hyper‐activ* )
 DocType=All document types; Language=All languages;
 #3 TS=((disrupt* near/3 (disorder* or behav*)) or (defian* near/3 (disorder* or behav* )))
 DocType=All document types; Language=All languages;
 #2 TS=((attention* or behav*) near/3 (defic* or dysfunc* or disorder* ))
 DocType=All document types; Language=All languages;
 #1 TS=(ADHD OR ADDH OR ADHS OR "AD/HD" OR hkd)
 DocType=All document types; Language=All languages;
 
 WorldCat (OCLC) (http://www.worldcat.org/)

All available years searched 28 June 2012 [no records] and 26 September 2013 [no records].

'kw:(ADHD OR hyper*) AND (amitriptyline OR amoxapine OR clomipramine OR clorimipramine OR chlorimipramine OR desipramine OR dosulepin OR dothiepin OR doxepin OR imipramine OR iprindole OR lofepramine OR maprotiline OR nortriptyline OR Opipramol OR Protriptyline OR trimipramine OR tricyclic*)' > 'Thesis/dissertation'
 
 ICTRP (http://apps.who.int/trialsearch/AdvSearch.aspx)

All available years searched 29 June 2012 [no records] and 27 September 2013 [no records].

Using Advanced search:
 CONDITION: ADHD OR hyper*
 INTERVENTION: amitriptyline OR amoxapine OR clomipramine OR clorimipramine OR chlorimipramine OR desipramine OR dosulepin OR dothiepin OR doxepin OR imipramine OR iprindole OR lofepramine OR maprotiline OR nortriptyline OR Opipramol OR Protriptyline OR trimipramine OR tricyclic*
 
 National Institute for Health Research Archive (http://www.nihr.ac.uk/Pages/NRRArchiveSearch.aspx)

Searched 29 June 2012 [no records]. Not searched in 2013 as this is an archive and no new content has been added.

adhd AND (amoxapine OR tricyclic OR amitriptyline OR amoxapine OR clomipramine OR clorimipramine OR chlorimipramine OR desipramine OR dosulepin OR dothiepin OR doxepin OR imipramine)

adhd AND (iprindole OR lofepramine OR maprotiline OR nortriptyline OR Opipramol OR Protriptyline OR trimipramine)
 
 ClinicalTrials.gov (http://clinicaltrials.gov/ct2/search/advanced)

All available years searched 29 June 2012 [no records] and 27 September 2013 [no records].

Using Advanced search:

Conditions: adhd OR hyperactivity OR hyperactive
 Interventions: amitriptyline OR amoxapine OR clomipramine OR clorimipramine OR chlorimipramine OR desipramine OR dosulepin OR dothiepin OR doxepin OR imipramine OR iprindole OR lofepramine OR maprotiline OR nortriptyline OR Opipramol OR Protriptyline

Conditions:adhd OR hyperactivity OR hyperactive
 Interventions: trimipramine OR tricyclic

Data and analyses

Comparison 1. TCA versus placebo.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 CGI (response rate)	3	125	Odds Ratio (M‐H, Random, 95% CI)	18.50 [6.29, 54.39]
1.1 Desipramine	2	103	Odds Ratio (M‐H, Random, 95% CI)	26.41 [7.41, 94.18]
1.2 Nortriptyline	1	22	Odds Ratio (M‐H, Random, 95% CI)	7.88 [1.10, 56.12]
2 Efficacy for ADHD symptoms (parents using a rating scale)	2		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
2.1 Desipramine	2	99	Std. Mean Difference (IV, Random, 95% CI)	‐1.42 [‐1.99, ‐0.85]
3 Efficacy for ADHD symptoms (parents using a rating scale) (response rate)	1		Odds Ratio (M‐H, Random, 95% CI)	Totals not selected
3.1 Desipramine	1		Odds Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
4 Efficacy for ADHD symptoms (teachers using a rating scale)	2		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
4.1 Desipramine	2	89	Std. Mean Difference (IV, Random, 95% CI)	‐0.97 [‐1.66, ‐0.28]
5 Efficacy for ADHD symptoms (parents using VAS)	1		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
5.1 Desipramine	1	68	Std. Mean Difference (IV, Random, 95% CI)	‐5.29 [‐6.32, ‐4.26]
6 CGI (% of change from baseline)	1		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
6.1 Desipramine	1	62	Std. Mean Difference (IV, Random, 95% CI)	‐1.27 [‐1.82, ‐0.72]
7 Child general behaviour (CBCL ‐ hyperactive subscale)	1		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
7.1 Desipramine	1	46	Std. Mean Difference (IV, Random, 95% CI)	‐2.05 [‐2.77, ‐1.32]
8 Child general behaviour (CBCL ‐ nervous / overactive subscale)	1		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
8.1 Desipramine	1	46	Std. Mean Difference (IV, Random, 95% CI)	‐7.88 [‐9.66, ‐6.10]
9 Child general behaviour (CBCL ‐ unpopular subscale)	1		Mean Difference (IV, Random, 95% CI)	Subtotals only
9.1 Desipramine	1	16	Mean Difference (IV, Random, 95% CI)	‐5.40 [‐9.26, ‐1.54]
10 Child general behaviour (CBCL ‐ anxious subscale)	1		Mean Difference (IV, Random, 95% CI)	Subtotals only
10.1 Desipramine	1	16	Mean Difference (IV, Random, 95% CI)	‐4.90 [‐9.82, 0.02]
11 Child general behaviour (CBCL ‐ obsessive compulsive subscale)	1		Mean Difference (IV, Random, 95% CI)	Subtotals only
11.1 Desipramine	1	16	Mean Difference (IV, Random, 95% CI)	‐6.50 [‐13.20, 0.20]
12 Depressive symptoms	1		Std. Mean Difference (IV, Random, 95% CI)	Totals not selected
12.1 Desipramine	1		Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
13 CGI ‐ ODD symptoms	1		Odds Ratio (M‐H, Random, 95% CI)	Totals not selected
13.1 Nortriptyline	1		Odds Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
14 All‐cause treatment discontinuation	3		Risk Difference (M‐H, Random, 95% CI)	Subtotals only
14.1 Desipramine	3	134	Risk Difference (M‐H, Random, 95% CI)	‐0.10 [‐0.25, 0.04]

Comparison 2. Desipramine versus clonidine.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Efficacy for ADHD symptoms (parent‐rated symptoms using VAS)	1		Std. Mean Difference (IV, Random, 95% CI)	Totals not selected
2 Child general behaviour (CBCL ‐ hyperactive subscale)	1		Std. Mean Difference (IV, Random, 95% CI)	Totals not selected
3 Child general behaviour (CBCL ‐ nervous / overactive subscale)	1		Std. Mean Difference (IV, Random, 95% CI)	Totals not selected
4 Child general behaviour (CBCL ‐ unpopular subscale)	1		Std. Mean Difference (IV, Random, 95% CI)	Totals not selected
5 Child general behaviour (CBCL ‐ anxious subscale)	1		Std. Mean Difference (IV, Random, 95% CI)	Totals not selected
6 Child general behaviour (CBCL ‐ obsessive compulsive subscale)	1		Std. Mean Difference (IV, Random, 95% CI)	Totals not selected

Comparison 3. TCA versus placebo: sensitivity analysis 1.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Efficacy for ADHD symptoms (parents using a rating scale)	1		Std. Mean Difference (IV, Random, 95% CI)	Totals not selected
1.1 Desipramine	1		Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
2 Efficacy for ADHD symptoms (parents using a rating scale) (response rate)	1		Odds Ratio (M‐H, Random, 95% CI)	Totals not selected
2.1 Desipramine	1		Odds Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
3 CGI	1		Std. Mean Difference (IV, Random, 95% CI)	Totals not selected
3.1 Desipramine	1		Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
4 CGI (response rate)	1		Odds Ratio (M‐H, Random, 95% CI)	Totals not selected
4.1 Desipramine	1		Odds Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
5 All‐cause treatment discontinuation	1		Risk Difference (M‐H, Random, 95% CI)	Totals not selected
5.1 Desipramine	1		Risk Difference (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]

Comparison 4. TCA versus placebo: sensitivity analysis 2.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Efficacy for ADHD symptoms (parents using a rating scale)	2		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
1.1 Desipramine	2	99	Std. Mean Difference (IV, Random, 95% CI)	‐1.42 [‐1.99, ‐0.85]
2 Efficacy for ADHD symptoms (parents using a rating scale) (response rate)	1		Odds Ratio (M‐H, Random, 95% CI)	Subtotals only
2.1 Desipramine	1	41	Odds Ratio (M‐H, Random, 95% CI)	60.80 [6.42, 575.51]
3 Efficacy for ADHD symptoms (teachers using a rating scale)	2		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
3.1 Desipramine	2	89	Std. Mean Difference (IV, Random, 95% CI)	‐0.97 [‐1.66, ‐0.28]
4 Efficacy for ADHD symptoms (parents using VAS)	1		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
4.1 Desipramine	1	68	Std. Mean Difference (IV, Random, 95% CI)	‐5.29 [‐6.32, ‐4.26]
5 CGI	1		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
5.1 Desipramine	1	62	Std. Mean Difference (IV, Random, 95% CI)	‐1.27 [‐1.82, ‐0.72]
6 CGI (response rate)	2		Odds Ratio (M‐H, Random, 95% CI)	Subtotals only
6.1 Desipramine	2	103	Odds Ratio (M‐H, Random, 95% CI)	26.41 [7.41, 94.18]
7 Child general behaviour (CBCL ‐ hyperactive subscale)	1		Std. Mean Difference (IV, Random, 95% CI)	Totals not selected
7.1 Desipramine	1		Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
8 Child general behaviour (CBCL ‐ nervous / overactive subscale)	1		Std. Mean Difference (IV, Random, 95% CI)	Totals not selected
8.1 Desipramine	1		Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
9 Child general behaviour (CBCL ‐ unpopular subscale)	1		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
9.1 Desipramine	1	16	Std. Mean Difference (IV, Random, 95% CI)	‐1.30 [‐2.40, ‐0.19]
10 Child general behaviour (CBCL ‐ anxious subscale)	1		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
10.1 Desipramine	1	16	Std. Mean Difference (IV, Random, 95% CI)	‐0.92 [‐1.97, 0.12]
11 Child general behaviour (CBCL ‐ obsessive compulsive subscale)	1		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
11.1 Desipramine	1	16	Std. Mean Difference (IV, Random, 95% CI)	‐0.90 [‐1.94, 0.14]
12 Depressive symptoms	1		Std. Mean Difference (IV, Random, 95% CI)	Totals not selected
12.1 Desipramine	1		Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
13 All‐cause treatment discontinuation	3		Risk Difference (M‐H, Random, 95% CI)	Subtotals only
13.1 Desipramine	3	134	Risk Difference (M‐H, Random, 95% CI)	‐0.10 [‐0.25, 0.04]

4.2. Analysis.

Comparison 4 TCA versus placebo: sensitivity analysis 2, Outcome 2 Efficacy for ADHD symptoms (parents using a rating scale) (response rate).

4.3. Analysis.

Comparison 4 TCA versus placebo: sensitivity analysis 2, Outcome 3 Efficacy for ADHD symptoms (teachers using a rating scale).

4.4. Analysis.

Comparison 4 TCA versus placebo: sensitivity analysis 2, Outcome 4 Efficacy for ADHD symptoms (parents using VAS).

4.5. Analysis.

Comparison 4 TCA versus placebo: sensitivity analysis 2, Outcome 5 CGI.

4.6. Analysis.

Comparison 4 TCA versus placebo: sensitivity analysis 2, Outcome 6 CGI (response rate).

4.7. Analysis.

Comparison 4 TCA versus placebo: sensitivity analysis 2, Outcome 7 Child general behaviour (CBCL ‐ hyperactive subscale).

4.8. Analysis.

Comparison 4 TCA versus placebo: sensitivity analysis 2, Outcome 8 Child general behaviour (CBCL ‐ nervous / overactive subscale).

4.9. Analysis.

Comparison 4 TCA versus placebo: sensitivity analysis 2, Outcome 9 Child general behaviour (CBCL ‐ unpopular subscale).

4.10. Analysis.

Comparison 4 TCA versus placebo: sensitivity analysis 2, Outcome 10 Child general behaviour (CBCL ‐ anxious subscale).

4.11. Analysis.

Comparison 4 TCA versus placebo: sensitivity analysis 2, Outcome 11 Child general behaviour (CBCL ‐ obsessive compulsive subscale).

4.12. Analysis.

Comparison 4 TCA versus placebo: sensitivity analysis 2, Outcome 12 Depressive symptoms.

Comparison 5. TCA versus placebo: sensitivity analysis 3.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Efficacy for ADHD symptoms (parents using a rating scale)	2		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
1.1 Desipramine	2	99	Std. Mean Difference (IV, Random, 95% CI)	‐1.42 [‐1.99, ‐0.85]
2 Efficacy for ADHD symptoms (parents using a rating scale) (response rate)	1		Odds Ratio (M‐H, Random, 95% CI)	Subtotals only
2.1 Desipramine	1	41	Odds Ratio (M‐H, Random, 95% CI)	60.80 [6.42, 575.51]
3 Efficacy for ADHD symptoms (teachers using a rating scale)	2		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
3.1 Desipramine	2	89	Std. Mean Difference (IV, Random, 95% CI)	‐0.97 [‐1.66, ‐0.28]
4 Efficacy for ADHD symptoms (parents using VAS)	1		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
4.1 Desipramine	1	68	Std. Mean Difference (IV, Random, 95% CI)	‐5.29 [‐6.32, ‐4.26]
5 CGI	1		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
5.1 Desipramine	1	62	Std. Mean Difference (IV, Random, 95% CI)	‐1.27 [‐1.82, ‐0.72]
6 CGI (response rate)	2		Odds Ratio (M‐H, Random, 95% CI)	Subtotals only
6.1 Desipramine	2	103	Odds Ratio (M‐H, Random, 95% CI)	26.41 [7.41, 94.18]
7 Child general behaviour (CBCL ‐ hyperactive subscale)	1		Std. Mean Difference (IV, Random, 95% CI)	Totals not selected
7.1 Desipramine	1		Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
8 Child general behaviour (CBCL ‐ nervous / overactive subscale)	1		Std. Mean Difference (IV, Random, 95% CI)	Totals not selected
8.1 Desipramine	1		Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
9 Child general behaviour (CBCL ‐ unpopular subscale)	1		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
9.1 Desipramine	1	16	Std. Mean Difference (IV, Random, 95% CI)	‐1.30 [‐2.40, ‐0.19]
10 Child general behaviour (CBCL ‐ anxious subscale)	1		Mean Difference (IV, Random, 95% CI)	Subtotals only
10.1 Desipramine	1	16	Mean Difference (IV, Random, 95% CI)	‐4.90 [‐9.82, 0.02]
11 Child general behaviour (CBCL ‐ obsessive compulsive subscale)	1		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
11.1 Desipramine	1	16	Std. Mean Difference (IV, Random, 95% CI)	‐0.90 [‐1.94, 0.14]
12 Depressive symptoms	1		Std. Mean Difference (IV, Random, 95% CI)	Totals not selected
12.1 Desipramine	1		Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
13 All‐cause treatment discontinuation	3		Risk Difference (M‐H, Random, 95% CI)	Subtotals only
13.1 Desipramine	3	134	Risk Difference (M‐H, Random, 95% CI)	‐0.10 [‐0.25, 0.04]

5.2. Analysis.

Comparison 5 TCA versus placebo: sensitivity analysis 3, Outcome 2 Efficacy for ADHD symptoms (parents using a rating scale) (response rate).

5.3. Analysis.

Comparison 5 TCA versus placebo: sensitivity analysis 3, Outcome 3 Efficacy for ADHD symptoms (teachers using a rating scale).

5.4. Analysis.

Comparison 5 TCA versus placebo: sensitivity analysis 3, Outcome 4 Efficacy for ADHD symptoms (parents using VAS).

5.5. Analysis.

Comparison 5 TCA versus placebo: sensitivity analysis 3, Outcome 5 CGI.

5.6. Analysis.

Comparison 5 TCA versus placebo: sensitivity analysis 3, Outcome 6 CGI (response rate).

5.7. Analysis.

Comparison 5 TCA versus placebo: sensitivity analysis 3, Outcome 7 Child general behaviour (CBCL ‐ hyperactive subscale).

5.8. Analysis.

Comparison 5 TCA versus placebo: sensitivity analysis 3, Outcome 8 Child general behaviour (CBCL ‐ nervous / overactive subscale).

5.9. Analysis.

Comparison 5 TCA versus placebo: sensitivity analysis 3, Outcome 9 Child general behaviour (CBCL ‐ unpopular subscale).

5.10. Analysis.

Comparison 5 TCA versus placebo: sensitivity analysis 3, Outcome 10 Child general behaviour (CBCL ‐ anxious subscale).

5.11. Analysis.

Comparison 5 TCA versus placebo: sensitivity analysis 3, Outcome 11 Child general behaviour (CBCL ‐ obsessive compulsive subscale).

5.12. Analysis.

Comparison 5 TCA versus placebo: sensitivity analysis 3, Outcome 12 Depressive symptoms.

Comparison 6. TCA versus placebo: sensitivity analysis 4.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Efficacy for ADHD symptoms (parents using a rating scale)	2		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
1.1 Desipramine	2	99	Std. Mean Difference (IV, Random, 95% CI)	‐1.42 [‐1.99, ‐0.85]
2 Efficacy for ADHD symptoms (parents using a rating scale) (response rate)	1		Odds Ratio (M‐H, Random, 95% CI)	Subtotals only
2.1 Desipramine	1	41	Odds Ratio (M‐H, Random, 95% CI)	60.80 [6.42, 575.51]
3 Efficacy for ADHD symptoms (teachers using a rating scale)	2		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
3.1 Desipramine	2	89	Std. Mean Difference (IV, Random, 95% CI)	‐0.97 [‐1.66, ‐0.28]
4 CGI	1		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
4.1 Desipramine	1	62	Std. Mean Difference (IV, Random, 95% CI)	‐1.27 [‐1.82, ‐0.72]
5 CGI (response rate)	3	125	Odds Ratio (M‐H, Random, 95% CI)	18.50 [6.29, 54.39]
5.1 Desipramine	2	103	Odds Ratio (M‐H, Random, 95% CI)	26.41 [7.41, 94.18]
5.2 Nortriptyline	1	22	Odds Ratio (M‐H, Random, 95% CI)	7.88 [1.10, 56.12]
6 CGI ‐ ODD symptoms	1		Odds Ratio (M‐H, Random, 95% CI)	Totals not selected
6.1 Desipramine	0		Odds Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
6.2 Nortriptyline	1		Odds Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
7 Depressive symptoms	1		Std. Mean Difference (IV, Random, 95% CI)	Totals not selected
7.1 Desipramine	1		Std. Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
8 All‐cause treatment discontinuation	3		Risk Difference (M‐H, Random, 95% CI)	Subtotals only
8.1 Desipramine	3	134	Risk Difference (M‐H, Random, 95% CI)	‐0.10 [‐0.25, 0.04]

6.2. Analysis.

Comparison 6 TCA versus placebo: sensitivity analysis 4, Outcome 2 Efficacy for ADHD symptoms (parents using a rating scale) (response rate).

6.3. Analysis.

Comparison 6 TCA versus placebo: sensitivity analysis 4, Outcome 3 Efficacy for ADHD symptoms (teachers using a rating scale).

6.4. Analysis.

Comparison 6 TCA versus placebo: sensitivity analysis 4, Outcome 4 CGI.

6.5. Analysis.

Comparison 6 TCA versus placebo: sensitivity analysis 4, Outcome 5 CGI (response rate).

6.6. Analysis.

Comparison 6 TCA versus placebo: sensitivity analysis 4, Outcome 6 CGI ‐ ODD symptoms.

6.7. Analysis.

Comparison 6 TCA versus placebo: sensitivity analysis 4, Outcome 7 Depressive symptoms.

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Biederman 1989.

Methods	Double‐blind, placebo‐controlled trial of desipramine.
Participants	Children who met DSM‐III criteria for ADHD. Total = 62. Boys = 58. Girls = 4. Age range = 6 to 17 years.
Interventions	All drugs were discontinued for at least 1 week before entering the protocol. Desipramine or placebo dose of 4.6 mg/kg given in 2 doses daily. The trial was of 6 weeks duration. Compliance with treatment was monitored by weekly pill counts. No other psychotropic agents or formal psychological or behavioural therapy administered during the trial. At the end of the trial, placebo‐treated patients who did not improve were offered an open trial of desipramine. Only desipramine‐treated patients had end‐of‐treatment blood and ECG testing.
Outcomes	Conners' ADHD Rating Scales. CGI Scale. Continuous Performance Task (CPT). Paired Associated Learning Task (PALT). Children's Depression Inventory (CDI). Subjective Treatment Emergent Symptoms Scale (STESS). Blood pressure.
Notes	Only patients who completed at least 3 weeks of treatment were included in the analysis.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Patients were assigned randomly by a computer‐generated list.
Allocation concealment (selection bias)	Unclear risk	The authors did not describe how allocation concealment was achieved.
Blinding (performance bias and detection bias) Subjective efficacy outcomes	Low risk	Medication presented as identical‐appearing tablets. Clinical ratings were obtained under double‐blind conditions. Laboratory personnel were unaware of the patients' clinical status or response to treatment.
Blinding (performance bias and detection bias) All‐cause treatment discontinuation	Low risk	Clinical ratings were obtained under double‐blind conditions. Laboratory personnel were unaware of the patients' clinical status or response to treatment.
Incomplete outcome data (attrition bias) Subjective efficacy outcomes	Low risk	Incomplete outcome data addressed: "Of the 73 patients screened, 62 (85%, 31 in each treatment condition) completed at least 3 weeks of the trial protocol to provide useable data. The 11 patients who terminated prematurely (before week 3) were not included in the data analysis". All 62 participants were accounted for at the end of the trial".
Incomplete outcome data (attrition bias) All‐cause treatment discontinuation	Low risk	This trial outcome is not influenced by attrition bias.
Selective reporting (reporting bias)	Low risk	All expected outcomes reported.
Other bias	Unclear risk	The authors acknowledge that the research was supported, in part, by grants from Merrell‐Dow Pharmaceutical Company (manufacturer of Desipramine).

Donnelly 1986.

Methods	Double‐blind, placebo‐controlled trial of desipramine.
Participants	29 boys aged 6 to12 years meeting DSM‐III criteria for ADHD. Total = 29. Boys = 29. Girls = 0. Mean age = 8.8 years.
Interventions	19 of the 29 children were drug‐treatment naive and the remaining 10 were medication free for 2 weeks. Following baseline investigations, 17 boys received a 14‐day course of incremental doses of 3.4 mg/kg/day desipramine and 12 received placebo. Duration of trial was 2 weeks.
Outcomes	Conners' Abbreviated Rating Scale (CABRS). CPT. Buschke Test. Plasma desipramine levels. Plasma biochemistry levels for norepinephrine (NE). 3‐methoxy‐4‐hydroxyphenylglycol (MHPG). Urine biochemistry levels for NE & DA (dopamine) and their metabolites. Subjective Treatment Emergent Symptoms Scale (STESS). Blood pressure and heart rate.
Notes	The trial was conducted at a day hospital at the National Institute of Mental Health where the participants received trial evaluations.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	The authors did not describe how random sequence was generated.
Allocation concealment (selection bias)	Unclear risk	The authors did not describe how random sequence was kept concealed.
Blinding (performance bias and detection bias) Subjective efficacy outcomes	Low risk	Desipramine and placebo capsules had an identical appearance.
Blinding (performance bias and detection bias) All‐cause treatment discontinuation	Low risk	Desipramine and placebo capsules had an identical appearance.
Incomplete outcome data (attrition bias) Subjective efficacy outcomes	Low risk	No patient discontinued, thus there were no missing data.
Incomplete outcome data (attrition bias) All‐cause treatment discontinuation	Low risk	This endpoint was not affected by attrition bias.
Selective reporting (reporting bias)	Low risk	All expected outcomes reported.
Other bias	Low risk	The trial appears to be free of other sources of bias.

Garfinkel 1983.

Methods	Double‐blind, placebo‐controlled, cross‐over trial of methylphenidate, clomipramine, and desipramine.
Participants	Children with DSM‐III diagnosis of ADD. Total = 12. Boys = 12. Girls = 0. Age range = 6 to 12 years. Mean age = 7.3 years.
Interventions	Each phase of this trial lasted 3 weeks. After an initial 2‐week baseline observation phase, each medication was used in random sequence with 7‐day washout intervals between treatment trial. Participants received capsules containing either 5 mg methylphenidate; 25 mg desipramine, 25 mg clomipramine, or 100 mg lactose. These were gradually titrated to average daily dosages of 18 mg methylphenidate, and 85 mg (3.5 mg/kg/day) for clomipramine or desipramine.
Outcomes	CTRS. Werry‐Weiss‐Peters Activity Rating Scale. Kagan Matching Familiar Figures Test. Reitan Maze. Reitan Holes. Blood pressure and heart rate.
Notes	6 of the subjects were previously treated with stimulant medication.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Participants randomly selected by 4 x 4 Latin square.
Allocation concealment (selection bias)	Unclear risk	The authors did not describe how randomisation sequence was concealed.
Blinding (performance bias and detection bias) Subjective efficacy outcomes	Low risk	"Medication was added to lactose powder and placed in identically appearing gelatin capsules".
Blinding (performance bias and detection bias) All‐cause treatment discontinuation	Low risk	"Medication was added to lactose powder and placed in identically appearing gelatin capsules".
Incomplete outcome data (attrition bias) Subjective efficacy outcomes	Low risk	This trial was a cross‐over design and paired analysis trial groups were comparable. Nevertheless, no result could be included in the performed meta‐analyses.
Incomplete outcome data (attrition bias) All‐cause treatment discontinuation	Low risk	This outcome was not affected by attrition bias.
Selective reporting (reporting bias)	Low risk	All expected outcomes were reported.
Other bias	Unclear risk	The trial had a cross‐over design and the carry‐over effect analysis was not reported.

Prince 2000.

Methods	Prospective, 6‐week, open phase followed by 3‐week, double‐blind, placebo‐controlled discontinuation trial of effect of nortriptyline.
Participants	23 of 32 children with DSM diagnosis of ADHD completed the double‐blind discontinuation phase. Total = 23. Boys = 18. Girls = 5. Age Range = 6 to17 years.
Interventions	The subjects were recruited to participate in a 6‐week, open trial of nortriptyline, which was titrated up to 2 mg/kg/day (mean = 1.8 mg/kg/day). Responders to treatment (i.e. those who had either a 30% reduction in the ADHD Rating Scale or a score of 1 or 2 on the CGI scale) were then randomised into a 3‐week controlled, discontinuation phase. During the discontinuation phase, subjects either continued on their current dose of nortriptyline or were tapered to placebo under double‐blind condition. 12 subjects received nortriptyline and 11 received placebo.
Outcomes	DSM‐IV ADHD Rating Scale. CGI. Children's Depression Inventory (CDI). RCMAS. ODD Rating Scale. CPT. Stroop Test. Rey‐Osterrieth Complex Figure (ROCF).
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	The authors reported that the subjects were randomly allocated but no information was given about how the randomisation process was performed.
Allocation concealment (selection bias)	Unclear risk	The authors did not describe how the allocation process was concealed.
Blinding (performance bias and detection bias) Subjective efficacy outcomes	Unclear risk	The authors did not describe how blinding was performed.
Blinding (performance bias and detection bias) All‐cause treatment discontinuation	Unclear risk	The authors did not describe how blinding was performed.
Incomplete outcome data (attrition bias) Subjective efficacy outcomes	Low risk	Only 2 out of 23 patients discontinued the double‐blind phase of the trial.
Incomplete outcome data (attrition bias) All‐cause treatment discontinuation	Low risk	This trial outcome was not influenced by attrition bias.
Selective reporting (reporting bias)	Unclear risk	The number of patients discontinuing in each trial arm, as well as the reasons for dropping out, were not reported.
Other bias	Unclear risk	Only those who responded positively in the open phase were randomised in double‐blind, placebo‐controlled phase. We found some possible inconsistencies in the figures reported. It is reported that 12 patients receiving nortriptyline and 11 receiving placebo completed the trial. These figures were used to calculate the effect of the intervention on ADHD symptom severity. However, when this endpoint is reported as "response rate" it seems that the number of patients used in the analysis was 13 and 9 for the groups receiving nortriptyline and placebo, respectively. This trial was supported, in part, by a pharmaceutical firm.

Singer 1995.

Methods	Randomized, placebo‐controlled, cross‐over trial of clonidine and desipramine.
Participants	Children aged 7 to 14 years meeting DSM‐III‐R criteria for Tourette syndrome and ADHD. Total = 34. Boys = 31. Girls = 3. Mean age = 10.6 years.
Interventions	34 children with Tourette syndrome and ADHD were offered uniform‐appearing capsules and each capsule contained a fixed amount of medication or placebo as shown below: Clonidine: up to 0.05 mg, 4 times daily for 6 weeks (mean = 0.19 mg/day); Desipramine: up to 25 mg, 4 times daily for 6 weeks (mean = 97 mg/day); Placebo: 4 times daily for 6 weeks. Each patient was maintained at the highest dose that did not produce side effects. A research nurse contacted all families on a weekly basis to monitor compliance and potential side effects. The duration of the trial was 6 weeks. After treatment for 6 weeks, the patients were evaluated for outcome measures. After an additional 1‐week, drug‐free interval, medication for the second (or third) treatment was started.
Outcomes	CBCL. Gordon Diagnostic System. Clinical Evaluation of Language Function. Matching Familial Figures Test. Porteus Maze Test. Restricted Academic Test. Tourette Syndrome Severity Scale. Hopkins Motor or Vocal Scale. Yale Global Tic Severity Scale. Leyton Obsessional Inventory.
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not described.
Allocation concealment (selection bias)	Low risk	"All health care providers, raters, parents, and patients were blinded to drug conditions and only a biostatistician had knowledge of the code".
Blinding (performance bias and detection bias) Subjective efficacy outcomes	Low risk	"Medications prepared by the Johns Hopkins Hospital pharmacy and provided as uniform appearing capsules in numbered containers. All healthcare providers, raters, parents and children were blinded to drug conditions".
Blinding (performance bias and detection bias) All‐cause treatment discontinuation	Low risk	"Medications prepared by the Johns Hopkins Hospital pharmacy and provided as uniform appearing capsules in numbered containers. All healthcare providers, raters, parents and children were blinded to drug conditions".
Incomplete outcome data (attrition bias) Subjective efficacy outcomes	Low risk	Incomplete outcome data addressed: "Of the 37 patients who entered the study, three were withdrawn either during or immediately after the first treatment phase. One was removed for failure to comply with medication administration and two at the request of the parents".
Incomplete outcome data (attrition bias) All‐cause treatment discontinuation	Low risk	This trial outcome was not influenced by attrition bias.
Selective reporting (reporting bias)	High risk	Not all outcome data provided. The investigators reported only significant positive findings.
Other bias	Unclear risk	This was a cross‐over trial: "Statistical analysis corrected for the presence of confounding effects (order and/or drug x order interaction) to eliminate any potential carryover effects and variability at the conclusion of each cycle". This analysis showed the existence of a carry‐over effect.

Spencer 2002.

Methods	Double‐blind, placebo‐controlled trial of desipramine and placebo.
Participants	5 to 17 year‐old children and adolescents with DSM‐IV diagnoses of combined‐type ADHD and chronic motor tic disorder, chronic vocal tic disorder or Tourette disorder. Total = 41. Boys = 34. Girls = 7.
Interventions	6‐week duration. Administered desipramine at 3.5 mg/kg/day and placebo. Safety monitoring: ECG, heart rate, and blood pressure taken weekly. Serum desipramine samples taken at the end of week 6.
Outcomes	CGI Scale. Conners' ADHD Rating Scale. Yale Global Tic Severity Scale (YGTSS). CYBOCS. Children's Depression Inventory (CDI). RCMAS. Global Assessment of Functioning (GAF).
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"Pharmacy randomised participants using separate balanced randomisation".
Allocation concealment (selection bias)	Low risk	"Randomisation codes were kept in sealed envelopes in the medical records. Medication was given in identical‐appearing 25 mg capsules".
Blinding (performance bias and detection bias) Subjective efficacy outcomes	Low risk	Randomisation was blinded to all trial personnel other than pharmacy staff. No subject was taking psychoactive medication within 1 month of the baseline assessment, and no additional psychoactive medication was allowed in the trial.
Blinding (performance bias and detection bias) All‐cause treatment discontinuation	Low risk	Randomisation was blinded to all trial personnel other than pharmacy staff No subject was taking psychoactive medication within 1 month of the baseline assessment, and no additional psychoactive medication was allowed in the trial.
Incomplete outcome data (attrition bias) Subjective efficacy outcomes	Low risk	Incomplete outcome data addressed. All participants completed the trial and there were no treatment withdrawals.
Incomplete outcome data (attrition bias) All‐cause treatment discontinuation	Low risk	This trial outcome was not influenced by attrition bias.
Selective reporting (reporting bias)	Low risk	All expected outcomes reported.
Other bias	Low risk	Trial groups were balanced at baseline. The trial appears free of other biases.

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Carlson 1995	The study did not evaluate ADHD symptoms. Subjects had mixed presentation of ADHD and depression.
Gualtieri 1988	This was a pilot study of 3 days' duration. The main focus was on neuropsychological drug effects on 9 children with ADHD. The study did not assess the core symptoms of ADHD.
Guardiola 1999	The study assessed the effects of pharmacological treatment on cognition and attention but not the core symptoms of ADHD.
Krakowski 1965	Subjects were hyperactive children not formally diagnosed with ADHD. Some of the participants were under 6 years of age.
Overtoom 2003	This was a laboratory investigation aimed at identifying the active sites of ADHD medications.
Pataki 1993	The aim of the study did not include assessment of efficacy of these medications. The trial was on hospitalised children with ADHD and depression. It was designed to specifically assess the side effects of methylphenidate and desipramine in these hospitalised children.
Rapoport 1974	Subjects were hyperactive but not formally diagnosed with ADHD.
Tellechea 1991	The study assessed the effects of pharmacological treatment on cognition and attention but not the core symptoms of ADHD.
Werry 1980	Subjects were hyperactive but not formally diagnosed with ADHD.
Yellin 1978	Subjects were hyperactive but not formally diagnosed with ADHD.
Yepes 1977	Subjects were hyperactive children with disturbed behaviour but were not formally diagnosed with ADHD.

Differences between protocol and review

We made minor edits and additions to the Background section to reflect current clinical knowledge. We distinguished primary from secondary outcomes to reflect the current standard of CDPLPG reviews. Initially, we planned to only include clinical trials lasting at least five weeks, but discarded this inclusion criterion to allow for the inclusion of short‐term studies. We screened for double‐blind studies and included more comparisons, such as TCA versus active agent, and TCA versus different TCA, in order to broaden the scope and clinical utility of this Cochrane Review.

We planned to conduct subgroup analyses for: individual TCA (e.g. desipramine, clomipramine, imipramine, amitriptyline); age groups (younger than 12 years of age, 12 to 18 years of age); and dosage but we could only perform subgroup analyses for individual TCA because of a lack of variability for age and dosage.

Also, we planned to handle studies with multiple comparisons differently, depending on whether interventions were independent or correlated. When several independent comparisons were available (e.g. TCA + psychotherapy versus placebo + psychotherapy versus TCA + fake psychotherapy versus placebo + fake psychotherapy), we planned to include them in the meta‐analysis as independent studies (in the example above: the comparison TCA + psychotherapy versus placebo + psychotherapy would be considered study one, and TCA + fake psychotherapy versus placebo + fake psychotherapy would be considered study two). In studies with multiple and correlated interventions, e.g. desipramine 4 mg/kg/day versus desipramine 6 mg/kg/day versus placebo, we planned to combine desipramine groups into a single group and include it in the meta‐analysis as one single comparison. We did not apply any of these methods since there were no multiple arms issues when we performed the statistical analysis of this systematic review.

To enhance the robustness of our search, we searched additional databases (see Appendix 1). Dissertation Abstracts International was no longer available to us, so we used WorldCat as a alternative source of theses and dissertations. XC and CHS helped to complete the review.

Contributions of authors

JO reviewed titles and abstracts; conducted searches, quality assessment, data extraction and entry; and wrote the review.

Xavier Castells (XC) conducted quality assessment, data extraction and data entry, and wrote the review.

Umonoibalo P Ehimare (UPE) reviewed titles and abstracts.

Clare H Smith (CHS) reviewed titles and abstracts.

Sources of support

Internal sources

• Worcestershire Health and Care NHS Trust, UK.

  Salary

External sources

• Sheffield Children's Hospital Charity, UK.

Declarations of interest

John Otasowie received a research grant from Sheffield Children's Hospital Charity for this Cochrane Review. The charity paid expenses for retrieving articles and also paid for John to attend a Cochrane Group training day.
 Xavier Castells ‐ none known.
 Umonoibalo P Ehimare ‐ none known.
 Clare H Smith ‐ none known.

New