4 Evidence and interpretation

The Appraisal Committee (see Appendix A) considered evidence from a number of sources (see Appendix B).

4.1 Clinical effectiveness

4.1.1 The Assessment Group found a total of 64 randomised controlled trials that met the inclusion criteria for the systematic review. An additional trial (the Multimodal Treatment Study of Children with ADHD [MTA] study) that did not meet the inclusion criteria was also included. This study did not meet the inclusion criteria because it assessed 'medical management' rather than a specific drug, but its objectives were still relevant to this appraisal.

4.1.2 The clinical studies used a large number of different instruments to measure key outcomes, core symptoms, and/or quality of life. This makes comparisons across the different trials difficult.

Methylphenidate compared with placebo

4.1.2.1 A large proportion of the studies found by the Assessment Group included a comparison between methylphenidate and placebo. Most of these studies were considered in the previous appraisal of methylphenidate in ADHD. The results of the additional studies used for this review were consistent with those in the previous appraisal. The evidence from short-term randomised placebo-controlled trials suggests that methylphenidate is an effective treatment to reduce core symptoms of ADHD in children who continue to take the medication.

Methylphenidate immediate-release compared with modified-release

4.1.2.2 There were seven studies comparing modified-release formulations of methylphenidate given once daily with immediate-release formulations administered two or three times daily. Of these, four were crossover studies and three were parallel studies. Of the four crossover studies, three used a modified-release formulation that is not marketed in the UK. All but one of the comparisons between modified- and immediate-release formulations included a placebo control.

4.1.2.3 With the exception of two very small crossover studies, most placebo-controlled studies found both immediate-release and modified-release methylphenidate to be superior to placebo in improving one or more core outcomes (thereby demonstrating that the effectiveness of both had been measured). Most studies did not indicate statistically significant differences in terms of measures of effectiveness when comparing the immediate-release and modified-release formulations with each other.

4.1.2.4 One 8-week comparison between Concerta XL and immediate-release methylphenidate in 145 children aged 6–12 years found that the mean change from baseline in a parent-administered symptom rating scale (SNAP IV) was significantly greater in the group who received Concerta XL. However, this study was open-label and so should be interpreted with caution.

4.1.2.5 In general, no statistically significant differences in the incidence of adverse effects were detected between immediate-release and modified-release formulations apart from in one study in which there was higher incidence of headache in participants assigned to the modified-release formulation.

4.1.2.6 Overall, the Assessment Report concluded that there was little evidence of a difference in the effectiveness of immediate-release and modified-release formulations of methylphenidate.

Dexamfetamine compared with placebo

4.1.2.7 Eight of the studies included in the Assessment Report compared dexamfetamine with placebo, and one study compared amfetamine with placebo. Of these, one study did not report effectiveness endpoints (this was a study of adverse events only and compared dexamfetamine with chlorpromazine and hydroxyzine as well as placebo). The remaining studies found improvements in behavioural symptoms with dexamfetamine (or amfetamine) relative to placebo, one of which related to a modified-release formulation of dexamfetamine (which is not available in the UK). The quality of these studies was considered to be variable by the Assessment Group and most studies did not score well in the quality assessment.

Comparisons of dexamfetamine and methylphenidate

4.1.2.8 The Assessment Report included four crossover studies that directly compared dexamfetamine and methylphenidate. Two of the four studies included a placebo control.

4.1.2.9 One of the placebo-controlled studies was a three-way crossover comparison in which participants received methylphenidate, dexamfetamine and placebo for 3 weeks each. Results from two instruments used for rating behavioural problems (Conners' Teachers' Rating Scale and Conners' Parent Questionnaire) were presented graphically. For both instruments, differences between dexamfetamine and placebo, and methylphenidate and placebo were reported to be statistically significant. The authors concluded that both drugs were highly and equally efficacious, but noted that frequently one drug or the other was preferable for any individual child.

4.1.2.10 The other study with a placebo group was a five-way crossover comparison of immediate- and modified-release methylphenidate, modified-release dexamfetamine (Dexedrine spansule), pemoline, and placebo. However, this study has limited relevance to this appraisal because neither modified-release dexamfetamine nor pemoline are licensed for use in the UK, nor the formulation of modified-release methylphenidate used in this study.

4.1.2.11 The studies without a placebo group comprised one two-way crossover comparison of methylphenidate and dexamfetamine, and a three-way crossover comparison—the third arm being treatment with caffeine capsules. The two-way comparison reported that methylphenidate was statistically significantly more effective than dexamfetamine when assessed using teacher-rated scales, but not when assessed using parent-rated scales. This study compared methylphenidate at a dose defined as 'medium' dose by the Assessment Group (0.6 mg/kg/day) while the dose of dexamfetamine was defined as 'low' (0.3 mg/kg/day). For the three-way comparison, methylphenidate and dexamfetamine were not statistically significantly different from each other for any of the effectiveness outcomes (it appears that both were significantly more effective than caffeine).

4.1.2.12 It appears that no statistically significant differences with respect to adverse effects were found between dexamfetamine and methylphenidate.

Atomoxetine compared with placebo

4.1.2.13 Eight of the studies included in the Assessment Report compared atomoxetine with placebo. Six of these studies were efficacy studies over 6–9 weeks in which participants were randomised to atomoxetine or placebo. Three of the studies used a once-daily dosing regimen for atomoxetine; the other three used a twice-daily regimen. All six of these studies found statistically significant differences in favour of atomoxetine on measures of hyperactivity and clinical global impression. However, one study that compared three doses of atomoxetine found that differences in hyperactivity and clinical global impression did not reach statistical significance in the lowest dose group (0.5 mg/kg/day).

4.1.2.14 Adverse effects included an increased risk of loss of appetite and weight loss in the atomoxetine-treated groups in some of the studies. Atomoxetine did not appear to be associated with an increased risk of headache, stomach ache or insomnia.

4.1.2.15 The other two studies investigated atomoxetine withdrawal. In one study, a total of 416 children and adolescents who had responded to open-label treatment with atomoxetine were randomised to continued atomoxetine or placebo for 9 months under double-blind conditions. At 9 months, fewer participants in the atomoxetine group had relapsed (relapse is defined as a return to 90% of baseline symptom severity) than in the placebo group (22.3% vs 37.9%; p = 0.002). Fewer details of the other discontinuation study are available in the Assessment Report. The authors concluded that discontinuation of atomoxetine was well tolerated.

Atomoxetine compared with dexamfetamine and methylphenidate

4.1.2.16 There were no studies directly comparing atomoxetine and dexamfetamine. There were three studies comparing atomoxetine and methylphenidate, two of which were unpublished. Two of these studies were open-label and did not include a placebo control. Because of this lack of blinding, the results of both studies should be interpreted with caution.

4.1.2.17 The published study was a 10-week open-label comparison between immediate-release methylphenidate and atomoxetine in children aged 7–15 years who had previously responded favourably to methylphenidate. There was no placebo control. The group sizes were uneven; 184 patients were randomised to atomoxetine, and 44 to immediate-release methylphenidate. This study reported no difference between the two drugs for hyperactivity or clinical global impression. However, a finding of no difference on subjective outcomes is difficult to interpret in the absence of a placebo group because it cannot be certain that drug effects were successfully measured in either group.

4.1.2.18 One unpublished study was a randomised, double-blind, placebo-controlled study comparing atomoxetine with modified-release methylphenidate (OROS formulation, Concerta XL) during acute treatment for 6 weeks. The study population comprised 516 children aged 6–16 years with ADHD (atomoxetine, n = 222; modified-release methylphenidate, n = 220; and placebo, n = 74). Patients may or may not have received previous treatment with stimulants, but those who had previously had an inadequate response to stimulant treatment were excluded from the study. The primary endpoint was response rate (defined as a reduction of 40% or more in ADHD rating scale [ADHD-RS] total symptom score from baseline). The response rate was 45% in the atomoxetine group, 56% in the modified-release methylphenidate group, and 24% in the placebo group. The response rates for both drugs were statistically significantly different from placebo, and the response rate for atomoxetine compared with modified-release methylphenidate was statistically significantly different (p = 0.016). Results of subgroup analyses for previously treated and treatment-naive participants were presented, but these were not interpreted using appropriate statistical tests for identifying subgroup effects.

4.1.2.19 The other unpublished study was a 3-week, open-label comparison of the OROS formulation of modified-release methylphenidate (that is, Concerta XL) and atomoxetine in 1323 children aged between 6 and 12 years. Participants were randomised in a 2:1 ratio to methylphenidate or atomoxetine; those who were known to be non-responders to treatments indicated for ADHD were excluded. This study reported significantly greater symptom improvement with modified-release methylphenidate than with atomoxetine in the ADHD rating scale for hyperactivity. This study did not score well in the Assessment Group's quality assessment.

4.2 Cost effectiveness

4.2.1 Seven published studies were found; five of these were economic evaluations and two were quality of life studies. The Assessment Group developed a model to compare the cost effectiveness of different drug strategies. Three consultees included economic evaluations in their submissions.

4.2.2 The results of the published economic evaluations are difficult to compare due to the use of different outcome measures. All studies suffered from a lack of data, and none considered the long-term outcomes or adverse events associated with ADHD. Only one study incorporated utility values and this reported an incremental cost-effectiveness ratio (ICER) of £9200 per quality-adjusted life year (QALY) gained for immediate-release methylphenidate compared with placebo.

4.2.3 The Assessment Group developed a probabilistic cost–utility model to compare the use of the drugs under consideration, both alone and in combination with behavioural therapy for a cohort of children with ADHD aged 6 years. In the base-case analysis, a 1-year time horizon was used. A secondary analysis extended the time horizon to the point when the cohort reached 18 years of age using an estimate of the age-dependent decline of symptoms. The base-case analysis considered alternative strategies featuring three active treatments (atomoxetine, dexamfetamine and one of the methylphenidate formulations), followed by no treatment as the last in the sequence.

4.2.4 The decision by the Assessment Group to consider three drug strategies rather than one or two, was based on their finding that each active treatment was cost effective relative to no treatment. The Assessment Group therefore considered it reasonable to assume that it would always be cost effective to change to the next untried drug, rather than stopping treatment after the first or second drug is found to be ineffective or not tolerated. This analysis relies on modelling assumptions, two of which are that response to one drug is independent of the response to another, and that response and withdrawal rates for second- and third-line treatments are the same as those for first-line treatment.

4.2.5 Clinical response was measured as a score of 1 or 2 (much improved or improved) on the clinician-rated clinical global impression improvement (CGI-I) subscale. Withdrawal rates were based on all reported withdrawals within trials. As non-response was included as a reason for withdrawal in some trials, this resulted in some double counting of non-responders. Non-drug costs were based on a published study that obtained estimates of resource use from a panel of experts. Drug costs were based on the average dose of active medication taken from the trials used in the calculation of response rates. Utility values were based on a published poster which derived utilities from EQ-5D questionnaires completed by the parents of 142 children with ADHD in the UK. Utility values were 0.837 for responders and 0.773 for non-responders, regardless of treatment type.

4.2.6 In the base-case analysis, 19 relevant strategies were compared, including a no treatment option. All strategies were cost effective compared with no treatment, with ICERs falling below £7000 per QALY gained. Given the limited data used to inform response and withdrawal rates and the small differences in QALY gains generated, it is not possible to distinguish between the different strategies on the grounds of cost effectiveness.

4.2.7 The results of the model comparing different treatment strategies with no treatment were relatively robust to the sensitivity analyses undertaken, including the addition of behavioural therapy, the use of different definitions of response, and the use of alternative utility values. The results were also robust when the time horizon was extended beyond 1 year using an estimated rate of remission. However, due to the limited data available, the long-term model did not incorporate the possible long-term adverse effects and benefits of treatment.

4.2.8 The manufacturer of Concerta XL (Janssen-Cilag) submitted a cost–utility analysis in which Concerta XL is compared over a 1-year period with immediate-release methylphenidate, atomoxetine, Equasym XL, and behavioural therapy in children with severe ADHD. The model assumes that children whose condition fails to respond to first-line therapy, or who experience intolerable side effects within 1 month, are switched to second-line treatment with behavioural therapy, combination therapy (behavioural therapy plus first-line pharmacotherapy) or other drug treatment (methylphenidate for patients receiving first-line behavioural therapy, otherwise dexamfetamine). Children not responding to second-line therapy within 1 month are assumed to discontinue all treatment. Based on these assumptions, Concerta XL is associated with an ICER of £5000 per QALY gained compared with immediate-release methylphenidate and dominates Equasym XL, atomoxetine, and behavioural therapy (that is, it is associated with more QALYs and a net cost saving relative to these alternative treatments).

4.2.9 The manufacturer of Equasym XL (UCB Pharma) submitted a cost–utility analysis in which Equasym XL is compared with no treatment in children with severe ADHD unable to comply with twice-daily immediate-release methylphenidate, over a 1-year period. A secondary analysis also compared Equasym XL with twice-daily immediate-release methylphenidate. The model assumes that children whose condition fails to respond to Equasym XL or who experience intolerable side effects within the 42-day titration period progress to second-line treatment with dexamfetamine. Non-compliers are assumed to continue on treatment, but experience no health benefits. Of those who discontinue second-line therapy, 50% progress to behavioural therapy and experience health benefits, while 50% progress to no treatment and experience no benefits. In the base-case analysis, Equasym XL is associated with an ICER of £14,700 per QALY gained compared with no treatment. In the secondary analysis, the ICER for Equasym XL compared with immediate-release methylphenidate is £11,000 per QALY gained.

4.2.10 The manufacturer of atomoxetine (Eli Lilly) submitted a cost–utility analysis examining the addition of atomoxetine to an existing medical management strategy for ADHD, over a 1-year period. The existing strategy consists of either immediate- or modified-release methylphenidate first line, followed by dexamfetamine second line and finally no treatment, with atomoxetine being added as an option prior to methylphenidate. A number of subgroups are considered depending on prior treatment history (naive or previously exposed to methylphenidate) and on whether the use of stimulants is appropriate (that is, whether they are contraindicated or not). The model assumes that children responding to treatment can relapse in subsequent cycles, and that adverse events either resolve in these cycles or result in discontinuation of treatment. For the different subgroups considered, the ICERs for atomoxetine compared with the relevant medical management strategy ranged from £11,500 per QALY gained for stimulant-naive patients with contraindications to stimulants to £15,400 per QALY gained for stimulant-exposed patients who responded to stimulants.

4.2.11 To summarise, the results of the published economic evaluations are difficult to compare. All studies suffer from a lack of data, and none consider the long-term outcomes or adverse events associated with ADHD. The results of the Assessment Group model suggest that methylphenidate, dexamfetamine and atomoxetine are all cost-effective treatments for ADHD. However, given the limited data used to inform response and withdrawal rates and the small differences in benefits between different treatments, it is not possible to compare different drug strategies. All three manufacturers adopted different approaches to the estimation of treatment effectiveness and associated utility values. However, the models all generated ICERs falling below £20,000 per QALY gained.

4.3 Consideration of the evidence

4.3.1 The Committee reviewed the evidence available on the clinical and cost effectiveness of methylphenidate, atomoxetine and dexamfetamine, having considered evidence on the nature of ADHD and the value placed by users on the benefits of these drugs, from children and adolescents with ADHD, their parents and guardians, those who represent them, and clinical experts. The Committee was also mindful of the need to ensure that its advice took account of the efficient use of NHS resources.

4.3.2 The Committee considered the evidence on clinical effectiveness and concluded that methylphenidate, atomoxetine and dexamfetamine are effective in controlling the symptoms of ADHD relative to no treatment. While some studies had included direct comparisons of different drugs and formulations, in general they reported few differences in measures of effectiveness between the products. In some studies, there were statistically significant differences in measures of effectiveness between drugs (see paragraphs 4.1.2.4, 4.1.2.18 and 4.1.2.19), but these studies had methodological flaws. In particular, the Committee considered the double-blind, placebo-controlled comparison of atomoxetine and modified-release methylphenidate (see 4.1.2.18). They considered that the exclusion of patients who have previously failed to respond to stimulants could have biased the result of this and other clinical studies comparing atomoxetine with methylphenidate. The Committee was not persuaded that superiority of one drug over another had been established in these trials. Given the large variations across the trials in measures of efficacy, the variable reporting of adverse events, and the lack of long-term studies, the Committee was not able to differentiate between the drugs on the grounds of clinical effectiveness.

4.3.3 The Committee understood that the individual drugs are associated with different contraindications and precautions for use. These may greatly influence the selection of appropriate therapy in children and adolescents with comorbid conditions. For example, atomoxetine may be preferred to methylphenidate and dexamfetamine for children with coexistent tic disorders or Tourette's syndrome. The Committee accepted the importance of having a range of drug treatment options.

4.3.4 The Committee noted the potential difficulties created by multiple daily dosing. In particular, concerns were raised regarding compliance and the social stigma associated with taking medicine, the availability and willingness of schools and school staff to store and administer medicine, and the potential for drug diversion (where the medication is forwarded on to others for non-prescription uses). The Committee therefore acknowledged that there would be situations in which a single-dose regimen, which can be achieved with modified-release formulations, would be the preferred treatment approach. However, the Committee concluded that there could be circumstances in which the limited range of dosage strengths available in the modified-release formulations would make titration difficult and an immediate-release formulation would be preferable.

4.3.5 With regard specifically to the use of dexamfetamine, the Committee noted that the licensed indication is limited to refractory hyperkinetic states. The evidence from clinical trials is generally of poor quality and relatively few studies have been conducted in recent years. The Committee also noted the concerns of the clinical and patient experts that dexamfetamine has a greater potential for diversion and misuse than the other drugs under consideration. Because of these limitations, the Committee acknowledged that dexamfetamine was unlikely to be used as a first-line drug for the majority of children or adolescents with ADHD. However, it concluded that dexamfetamine should remain a treatment option for use in specific situations. The Committee expected that clinicians experienced in the management of ADHD would take into account these considerations when initiating drug treatment for a child or adolescent with ADHD.

4.3.6 The Committee carefully considered all of the evidence on cost effectiveness and concluded that all three drugs are cost effective relative to no drug treatment. It reviewed the Assessment Group's modelling approach and noted the differential cost effectiveness of adopting different drug treatment sequences according to this analysis. The Committee noted that some of the strategies considered in the cost effectiveness analysis might be unsuitable for some individuals because of considerations of adverse events, comorbidities and concordance with therapy. On this basis and given the limitations inherent in the models, the Committee was unable to draw conclusions on the relative cost effectiveness of different drug treatment strategies. Although each of the drugs being appraised is acceptably cost effective versus no treatment, the Committee understood that some treatment strategies might be more cost effective than others in individual patient circumstances.

4.3.7 The Committee noted the variation in the costs of the drugs and treatment regimens. It also noted that since the unit cost of a dose of atomoxetine is the same regardless of the strength, twice-daily dosing could double the cost of treatment with this drug. The Committee considered that for the majority of potential users, where there is a choice of more than one appropriate product on clinical grounds, the product with the lowest cost (taking into account the cost per dose and number of daily doses) should be prescribed.

4.3.8 Overall, the Committee concluded that there were a number of important factors to be taken into account when selecting a treatment for an individual child or adolescent with ADHD. These included consideration of concordance and compliance issues, particularly with respect to the timing of doses, and whether the individual has difficulties relating to the administration of doses during the day, for example at school. Other important considerations include previous adverse effects, comorbidities, and the preferences of patients and carers. All of these factors may influence the choice of product.

4.3.9 On the basis of evidence from experts, the Committee concluded that treatment should only be initiated by an appropriately qualified healthcare professional with expertise in ADHD and should be based on a comprehensive assessment and diagnosis. However, the Committee agreed that continued prescribing and monitoring of drug therapy may be performed by general practitioners, under shared care arrangements.