While the results of prior ADHD pharmacogenetic studies have been intriguing, many challenges must still be addressed. For example, the majority of ADHD pharmacogenetic studies published to date have examined response to methylphenidate. Study of genetic predictors of response for additional ADHD medication treatment options, including amphetamine preparations, atomoxetine and guanfacine, remains a significant gap in the literature.
Furthermore, the prior methylphenidate pharmacogenetic studies have been hampered by a number of limitations, and have often yielded inconsistent findings. Study design differences may partially account for the heterogeneity of findings. For example, in the case of the dopamine transporter studies, more consistent findings appear to be emerging from placebo-controlled, prospective studies of children with ADHD. However, most prior trials have depended on open-label or retrospective assessment, in which medication doses were not specified or were considerably lower than those used in community practice for optimal benefit.[23
] Since the effects of methylphenidate on ADHD symptoms often follow a linear dose-response curve,[27
] these lower doses might bias against finding significant treatment effects.
Current studies are also constrained by the type of outcome measures used, as many studies rely on simple dichotomous outcomes (e.g. responder vs. nonresponder), which have limited power to detect effects compared with analyses of quantitative measures. Correlations between multiple outcome measures in the same subjects are also known to be fairly weak, raising the question as to which outcome measure best defines positive response.[97
] In several cases, study results have differed depending on whether parents or teachers are the behavior-rating informants.[28
An additional critical methodological issue is the approach to defining genotypes for analysis. In order to minimize the potential for spurious findings and increased type I errors, investigators must limit their analyses to minimal genotype combinations. For some genes, the risk polymorphisms for ADHD are the less common variants (e.g. the 7-repeat allele of DRD4), while for other genes, such as the dopamine transporter SLC6A3, it is the more common variants that are associated with the disorder. For SLC6A3, the 10/10 and 10/9 genotypes are most common, and earlier studies combined these common genotypes. This practice assumed a dominant effect of either the SLC6A3 9 or 10 allele, but failed to test for a recessive effect of the 9/9 genotype. Alternative grouping of genotypes based on the presence of one or more SLC6A3 9-repeat alleles has led to different results. Future candidate gene studies would benefit from consensus on optimal strategies to define genotype groupings. Genotypes should not be lumped together when evidence of the dominance of one allele is lacking in previous pharmacogenetic studies.
Variation in sample size, composition and environmental exposures may also contribute to differences in ADHD pharmacogenetic study results. Modest sample sizes have limited statistical power to detect mild or moderate genetic effects. Another potential contributor to the observed discrepancies in study findings is that pharmacogenetic effects may vary in different ethnic and racial groups. This suggests that the genetic variants being studied may not be causing the effect observed, but instead may be in linkage disequilibrium with the actual functional genetic variants. In addition, previous investigations may have failed to identify consistent genetic effects due to differences in sample subtype composition, given evidence in some prior studies that certain genetic variants may influence response to medication in terms of hyperactive-impulsive symptoms[26
] or inattentive symptoms[24
] but not both domains. In addition, although ADHD pharmacogenetic studies to date have not evaluated interactions with additional environmental and toxicant exposures, prior evidence hints that such exposures may be important modifiers of genetic effects on medication response. For example, environmental exposures such as tobacco smoke may influence brain dopamine release by interacting with both catecholamine-related genetic variants[98
] and methylphenidate.[100
] If the suggested three-way interaction between catecholamine genes, methylphenidate and tobacco exposure is in fact at play, we would expect the measured associations between genotype and medication response to vary according to the tobacco exposure level of the different study populations.
Furthermore, failure to evaluate gene-gene interactions, which have received little attention in ADHD pharmacogenetic studies to date,[41
] may also be obscuring effects. Moreover, it is increasingly recognized that drug response is the result of a complex matrix of factors, rather than a single factor.[48
] As a result, experts have proposed that future pharmacogenetic studies shift their focus from individual genes to pathways encompassing genes for drug-metabolizing enzymes and transporters, as well as genes encoding drug targets and their downstream signals.[101
To address the limitations of previous studies, guiding principles to promote future ADHD pharmacogenetics research were proposed during the 2006 and 2008 ADHD Molecular Genetics Network (AMGN) meetings.[102
] These include the recommendation that studies utilize a methodologically rigorous pharmacological intervention, which typically means that investigations should be prospective, randomized and placebo-controlled. The AMGN also proposed that trials employ a full range of dose conditions, since dose-ranging and forced titration designs are more likely to elicit pharmacogenetic effects than flexible dosing. Further suggestions stressed the need for genotyping quality control, ideally including cross-laboratory and cross-method reliability checks. In addition, the AGMN recommended the recruitment of study samples large enough to evaluate effect modification by additional environmental exposures, as well as gene-gene interactions and/or a gene pathways approach.[101
] Finally, the group called for assessment of multiple outcomes, including both continuous and categorical outcomes examining both parent and teacher ratings, and proposed that functional outcomes and adverse events should be evaluated as well as symptom ratings.