Conducting stratified analyses based on MSDP provides great insight into the complex association between SLC6A2
and ADHD. Although pharmacological, imaging, and neuropsychological studies have extensively implicated the norepinephrine transporter in ADHD, genetic studies have shown a minimal association. Although associations have been reported, non-replication between studies has resulted in a lack of overall significance when a meta-analysis was conducted. 
Results presented here, and in an earlier report, 
support the view that the lack of replication between studies may be explained, at least in part, by the inherent clinical and etiological complexity of the disorders.
The association between MSDP and ADHD is one of the most investigated in the field of environmental psychiatric epidemiology. Although consistently replicated 
and high in magnitude, there is now relative consensus that this association has little causal significance 
and may instead be driven by other variables that are shared by the behavior of smoking during pregnancy in mothers and ADHD in their children. While environmental factors may play a role in this association, it is believed that genetic factors shared by mother and child play an important role in smoking during pregnancy in the former and ADHD in the latter. In this study, MSDP was used to index a subtype of ADHD with putatively more homogeneous genetic determinants shared within families of children with ADHD where mothers smoked during pregnancy. Consistent with this hypothesis, we have reported 
that children in this subgroup present a more severe clinical picture with greater behavioral problems and lower cognitive function, when compared to children whose mothers did not smoke during pregnancy, and that this difference in clinical phenotype is significant even when important environmental factors are controlled for. The results of the current study emphasize the genetic differences in these two subtypes. Polymorphisms (rs36021 and linked SNPs) are important genetic determinants of behavior, cognition, and treatment response in ADHD children whose mothers smoked during pregnancy, and who may represent a more homogeneous group of ADHD patients, as previously reported. 
In the subtype where mothers did not smoke during pregnancy, an association with a different region of the gene (towards the 3′ end of SLC6A2)
Given that the association between ADHD and rs36021 (and linked SNPs) is highly significant only in those children whose mothers smoked during pregnancy may suggest a true interaction between exposure to maternal smoking and carrying the risk allele(s) in the SLC6A2 gene. Indeed, the adverse consequences of in utero
exposure to the toxic effects of nicotine are well documented, from animal and human studies. 
MSDP is associated with pre- and peri-natal complications, deficits in cognitive development as well as long-term behavioral problems. Alternatively, but not exclusively, the etiology of smoking behavior and ADHD may involve closely related, but distinct pathways. Indeed, it is possible that the complex genetic background underlying smoking behaviors in mothers (which is transmitted in part to their children), interacts with risk alleles in SLC6A2
to increase the risk for ADHD in children. Under the latter scenario, MSDP may be considered as a phenotypic index used to select a subgroup of children with relatively more homogeneous genetic etiology.
Irrespective of the precise links between these pathways, this study strongly suggests that genetic variation in the SLC6A2
is an important factor in a more severe subtype of ADHD. If replicated in independent studies, this may represent an important step towards personalized medicine
in treating children with ADHD. 
Results of the present study are perfectly congruent with reports by Song et al
and Yang et al
but only in the group where mothers did not smoke during pregnancy. In this group, a significant over-transmission of the G
allele to the higher difference scores was observed in the quantitative FBAT analysis on the Conners’-T (). Most of this effect appears to arise from the restless-impulsive factor scores, observed only in the group of non-smoking mothers. It is noted that when treatment response was assessed using the CGI-Improvement scale, two previous studies, 
as well as the current study, did not find an association with 1287(G/A) (rs5569) ().
Several other previously-reported associations were replicated in the present study. Three studies had reported an association with rs3785143 and rs11568324. 
These markers are in complete LD with rs36021 (D
1; albeit with a low correlation coefficient, r2
; ), indicating that the 3 SNPs are in one haplotype block not separated by a recombination event. In the total sample, rs3785143 showed marginal association with ADHD, but a significant association with all CBCL dimensional scores (). No association was observed when stratified analyses were carried out. Similarly, no association was observed with rs11568324 despite the fact that it is in complete LD with rs36021. This is most likely a result of the low heterozygosity of these markers, which make them non-informative in the FBAT analysis (as indicated by the number of informative families in ). Two other previously-implicated SNPs, rs998424 and rs36017, showed marginal association with dimensions of ADHD in the sample where mothers did not smoke during pregnancy and the total sample, respectively.
Kim and colleagues 
reported an association between ADHD and a functional promoter SNP rs28386840 [-3081(A/T)] in several independent case-control studies. This association was not replicated in the current study, neither in the total sample, nor in the samples stratified by MSDP (, , ). The lack of association with ADHD was also reported in two other family-based studies. 
A study examining the association between this polymorphism and treatment response reported an association with CGI-improvement scores 
, where T
-allele carriers showed a better response to MPH treatment. In the current study, only a marginal association was observed with difference scores on the restless-impulsive subscale of the Conners’-T in the group where mothers smoked during pregnancy ().
In a previous report, 
we investigated the association between ADHD and the panel of 30 SNPs examined in the present study, and noted that a complex pattern of association emerged between SLC6A2
SNPs/haplotypes, ADHD subtypes and gender. Gender and subtype are considered two dimensions that might help in reducing genetic heterogeneity in the ADHD syndrome. Although these results helped explain some of the discrepancies noted among previous studies, stratification according to these dimensions did not yield as strong an association with SLC6A2
as the stratification based on MSDP, which may suggest that the latter is more pertinent for future efforts to map genes implicated in ADHD.
SNPs that showed the most significant association in this study (rs36021 and rs3785152, in particular) are within introns, opening up two possibilities. The first possibility is that these intronic variants are involved in gene regulation. The second is that these polymorphisms are not the causal mutation, but are in LD with a functional variant. Fine-mapping of the region is required to identify the causal mutation(s) followed by molecular analysis to determine if the mutation affects transcriptional regulation of the gene or structure and function of the protein.
While we conducted a large number of comparisons and some correction for multiple testing is warranted, it is important to note that when we correct for multiple testing in relation to our primary hypothesis, that is association between SLC6A2
and ADHD in children stratified according to MSDP, the primary result of association (Z
0.0002) with rs36021 remains significant even if we apply the overly stringent Bonferroni correction (30 SNPs times two exposure strata, p
0.002). In addition, the widespread exploratory associations that are observed with behaviors relevant to ADHD measured by different observers (parents, teachers, and research staff) and in different settings (school, home, clinic) with rs36021 suggest that these associations are unlikely to be chance findings. We believe that this considerable consistency of results strengthens the overall credibility of the primary results and help to understand how genetic vulnerability to ADHD is mediated through the traits and endophenotypes underlying this disorder.
To our knowledge, this is the largest study (among family-based and case-control studies) testing the association between ADHD and SLC6A2
, with such detailed genotype and phenotype characterization. While collaboration between multiple research groups in large consortia is vital for genetic studies of ADHD, it has been shown that a significant amount of heterogeneity can be introduced in multicenter collaborative studies because of divergent clinical and evaluation practices. 
This underscores the value of the current study where a relatively large sample has been collected at a single center using a highly standardized approach. It is also the largest study worldwide to use a double-blind, placebo-controlled design for evaluation of treatment response, combining extensive evaluation of executive function and behavioral domains, with genetic and environmental data. Nonetheless, these results must be considered exploratory and independent replication is awaited.
If confirmed in independent studies, these results will help to disentangle the complex etiological pathways of ADHD. In the long term, this would very likely lead to development of therapeutics targeting specific biochemical pathways in specific sub-groups of children with ADHD.