A total of 109 participants were enrolled. Two boys, one from each TaqIA genetic group, had a risperidone concentration less than 0.5λng/ml and were excluded from further analyses because of suspected medication nonadherence. In the remaining 107 participants (87% males), although externalizing disorders were the most prevalent psychiatric conditions, comorbidity was common (). Risperidone was administered to target irritability and aggression in 76% (n
=81) of the sample, tic disorder in 10% (n
=11), severe impulsivity in 8% (n
=8), sleep problems in 3% (n
=3), obsessive compulsive disorder in 2% (n
=2), and mood symptoms in one child. The dose prescribed is consistent with published trials [28
]. In addition to risperidone, participants were most often taking psychostimulants (65%), α2-agonists (52%), and SSRIs (32%), concurrently.
The median prolactin level was 18.7λng/ml (interquartile range: 13.4–28.5) with 50% of the participants exhibiting hyperprolactinemia as defined earlier. With a cutoff of 25.7λng/ml, regardless of sex, age, or pubertal status [29
], 34% (n
=36) of the sample had hyperprolactinemia.
Using multiple linear regression, we tested whether stage of sexual development, sex, SSRI treatment, dose of psychostimulants and risperidone per kilogram of body weight, or duration of risperidone treatment were associated with prolactin concentration in the entire sample (n=107). We found that the dose of risperidone (partial R2=0.1, P=0.0009) and Tanner stage (partial R2=0.07, P<0.003) positively predicted prolactin concentration, whereas the dose of psychostimulants (partial R2=0.04, P<0.03) was negatively associated with it. None of the other covariates were significant. When age was substituted for Tanner stage, comparable results were found with age accounting for 3.5% of the variance in prolactin concentration (partial R2=0.035, P<0.04). When risperidone concentration was substituted for the dose, it accounted for 39% (P<0.0001) of the variance in prolactin concentration with Tanner stage (or age) and the dose of psychostimulants becoming nonsignificant. Similar results were found when these analyses were restricted to males (females were a small proportion of the total sample).
As illustrated in , of 99 participants queried, 14% reported AEs potentially related to hyperprolactinemia. Of the 11 queried girls, five (45%) reported at least one AE compared with 10% (n=9) of boys (Fisher’s Exact P=0.008). Although most participants did not report any AE potentially related to hyperprolactinemia, 23% (n=12) of those with hyperprolactinemia endorsed at least one AE compared with 4% (n=2) of those without (Fisher’s Exact P=0.01). In fact, all the males and the three girls who reported at least one AE had hyperprolactinemia ().
Owing to racial/ethnic differences in the distribution of the different DRD2
], we restricted the genetic analyses to nonHispanic Caucasians who formed 84% (n
=90) of the sample. The genotype frequencies were as follows: TaqIA A2A2 63% (n
=57), A1A2 33% (n
=30), and A1A1 3% (n
=3); C957T CC 24% (n
=22), CT 44% (n
=40), and TT 31% (n
=28); -141 Ins/Del
Ins/Ins 77% (n
=69), Ins/Del 22% (n
=20), and Del/Del 1% (n
=1); and A-241G AA 87% (n
=78), AG 13% (n
=12), and GG 0%. All were in Hardy—Weinberg equilibrium. Using the Haploview program [30
] (), TaqIA and -141C Ins/Del
were found to be incomplete, but AQ3nonsignificant (D’1, R2
=0.04, LOD AQ4score 0.79), linkage disequilibrium (LD). Significant LD was detected between C957T and TaqIA (D’0.76, R2
=0.17, LOD score 3.39) and C957T and -141C Ins/Del (D’0.86, R2
=0.12, LOD score 2.69). A-241G was not in LD with any of the three other variants.
Fig. 1 Pairwise linkage disequilibrium (LD) analysis across the four DRD2 variants. Light grey: LD (moderate-to-high D’) between C957T and TaqIA and between C957T and -141C Ins/Del. White: almost no significant LD between the variants. Empty white: complete, (more ...)
As shown in , there were no demographic or clinical differences between the two TaqIA genetic groups except for patients with the A2A2 genotype being more likely to have attention deficit hyperactivity disorder (P<0.006).
Prolactin concentration was higher in the TaqIA A1 allele carriers (P
=0.03). Furthermore, whether defined based on the laboratory upper normal limit for males and females or using a single cutoff of prolactin concentration greater than 25.7λng/ml [29
], the rate of hyperprolactinemia was 1.5–2 times higher in the A1 allele carriers compared with participants with the A2A2 genotype (P
≤0.04). We found similar results when we matched the two genetic groups on the daily dose of psychostimulant per kilogram of body weight. In fact, the median prolactin concentration was 24.0λng/ml in the A1 allele carriers and 16.2 in the A2A2 group (Wilcoxon’s signed rank S
=0.04) and the rate of hyperprolactinemia was 65% (n
=22) in the A1 allele carriers versus 34% (n
=11) in the A2A2 group (McNemar’s S
=7.1, d.f.=1, P
We then used multiple linear regression to predict prolactin concentration adjusting for the TaqIA genetic group as well as the significant covariates identified earlier (i.e. age or Tanner stage, the dose of risperidone or its serum concentration, and the dose of psychostimulants). As expected, the daily dose of risperidone per kilogram of body weight increased and that of psychostimulants decreased prolactin concentration. We also found a trend for prolactin to increase with age (F(1,85)=2.7, P=0.1) and with the TaqIA A1 allele carrier state [adjusted least squares means of log prolactin 3.2λng/ml in A1 allele carriers versus 2.9 in noncarriers, β=0.24, F(1,85)=3.2, P<0.08]. This genetic effect accounted for 2.8% of the variance in log prolactin concentration (ΔR2=0.028). Similar results were obtained when the stage of sexual development, instead of age, was included in the model. When the risperidone concentration was substituted for the oral dose, the TaqIA effect became significant (β=0.24, F(1,78)=4.3, P=0.04, ΔR2=0.030) while the opposite was true for the dose of psychostimulants (F(1,78)=2.5, P=0.1). However, neither age nor Tanner stage significantly contributed to the model (P>0.4). Finally, we used logistic regression to compare the rate of hyperprolactinemia across the two TaqIA genetic groups while controlling for the concentration of risperidone and the dose of psychostimulants per kilogram of body weight. We found that the adjusted odds of developing hyperprolactinemia for A1 allele carriers were 3.1 times (OR: 95% CI: 1.1–9.7, Wald χ2=4.0, P<0.05) greater than for participants with the A2A2 genotype. The risperidone concentration also increased the probability of hyperprolactinemia (Wald χ2=14.8, P<0.0001), whereas the dose of psychostimulants had no effect (P=0.3). When the analyses were restricted to boys, the pattern of the results remained unchanged.
Next, we explored the effects of the -141C Ins/Del, C957T, or A-241G variants on prolactin concentration using multiple linear regression and controlling for the serum concentration of risperidone and the dose of psychostimulants per kilogram of body weight. Although the former continued to have a strong positive effect on prolactin, neither psychostimulants nor the -141C Ins/Del variant had a significant effect (P>0.1 and P=0.6, respectively). When C957T was substituted for -141C Ins/Del, we again found a strong effect of risperidone concentration (P<0.001) but no effect of psychostimulants (P>0.2). In addition, there was a trend for the three genotypes to predict prolactin, with the CT genotype having the highest concentration (F(2,77)=2.6, P=0.08, ΔR2=0.037). With a recessive mode of transmission, C957T had no significant effect (P=0.4). In contrast, after controlling for the serum concentration of risperidone (F(1,78)=52.2, P<0.0001) and the dose of psychostimulants (F(1,78)=5.6, P=0.02), having the AA genotype of the A-241G variant was associated with a strong protective effect (β=-0.53, F(1,78)=10.0, P=0.002). This genetic effect accounted for 6.6% of the variance in prolactin concentration (ΔR2=0.066). When both TaqIA and A-241G genetic groups were introduced into the regression model predicting prolactin, both independently contributed to the model [β=0.23, F(1,77)=4.1, P<0.05, ΔR2=0.026 and β=-0.51, F(1,77)=9.7, P<0.003, ΔR2=0.062, respectively] in addition to risperidone serum concentration and the dose of psychostimulants. Finally, after controlling for the concentration of risperidone and dose of psychostimulants, we found a synergistic effect of the TaqIA and A-241G variants on prolactin concentration (F(3,76)=5.1, P=0.003) (). In fact, compared with those participants with the TaqIA A2A2/-241AA joint genotype, participants with the A1 allele/AA genotype experienced a mean increase in prolactin (log transformed) of 7% (n=28, t=1.5, P=0.1) whereas those with the A2A2/-241G allele genotype experienced a 14% mean increase in prolactin (n=7, t=1.8, P=0.07). In participants with the A1/-241G alleles joint genotype, the mean prolactin concentration was 31% higher (n=5, t=3.6, P=0.0006).
Adjusted least squares mean of prolactin concentration (log transformed) as a function of the TaqIA and A-241G joint genotypes of the dopamine D2 receptor gene in children and adolescents in long-term treatment with risperidone.
Finally, after controlling for hyperprolactinemia status (Wald χ2=4.1, P=0.04), the odds of developing at least one AE potentially related to hyperprolactinemia was 4.4 times greater in TaqIA A1 allele carriers compared with A2A2 participants (OR: 95% CI: 1.1–76.0, Wald χ2=4.0, P<0.05). However, although hyperprolactinemia status continued to increase the risk for AE (Wald χ2=5.0, P<0.03), the A-241G variants did not affect this risk, likely because of the small number of A-241G allele carriers (n=11). Similar analyses were not conducted with the other variants, owing to their lack of effect on prolactin concentration, or with the joint genotypes because of the small size of the different genetic subgroups.