Abstract
Aims:
The primary aim of this study was to determine whether CYP2D6 phenotypes are associated with atomoxetine treatment outcomes in children and adolescents with attention-deficit hyperactivity disorder.
Patients and Methods:
Patients 21 years and younger with a known CYP2D6 genotype and a prescription for atomoxetine were included in this electronic health record review. The primary outcome was defined as discontinuation of atomoxetine due to a lack of efficacy or due to toxicity. CYP2D6 activity scores were adjusted to account for phenoconversion if patients were concomitantly taking a CYP2D6 inhibitor medication.
Results:
Of 108 patients that met inclusion criteria, including 13 poor, 30 intermediate, 61 normal, and 4 ultra-rapid CYP2D6 metabolizers after adjusting for phenoconversion. Normal/ultrarapid metabolizers were significantly more likely to discontinue atomoxetine due to lack of efficacy (subdistribution hazard ratio [sHR] = 1.9, p = 0.04), while poor/intermediate metabolizers were significantly more likely to stop due to toxicity (sHR 0.4, p = 0.03). A significant nonlinear effect of adjusted activity score on likelihood of discontinuation due to lack of efficacy (p < 0.001) or toxicity (p < 0.001) was identified, where higher CYP2D6 activity scores were associated with more discontinuation due to lack of efficacy and lower activity scores were associated with more discontinuation due to toxicity. When CYP2D6 genotype was available prior to prescribing atomoxetine in intermediate/poor metabolizers, lower median initial doses were observed (0.5 mg/kg/day) compared to those who underwent CYP2D6 genotyping after the initial atomoxetine prescription (0.8 mg/kg/day).
Conclusion:
Patients with normal or increased CYP2D6 phenotypes may be more likely to discontinue atomoxetine due to a lack of efficacy, while those with reduced function CYP2D6 phenotypes may be more likely to discontinue atomoxetine due to toxicity.
Introduction
Attention-deficit hyperactivity disorder (ADHD) is one of the most common neuropsychiatric conditions affecting children. Atomoxetine was the first non-stimulant medication to be approved for the treatment of ADHD in the United States (Savill et al., 2015), with its mechanism of action related to its selective inhibition of presynaptic norepinephrine reuptake in the prefrontal cortex. The noradrenergic system is important for attention, learning, and memory, and the increase in extracellular norepinephrine levels that result from treatment with atomoxetine is believed to help treat ADHD symptoms (Ledbetter, 2006). Following its approval in 2002, atomoxetine was the third most commonly prescribed medication for the treatment of ADHD after the first-line stimulant medications methylphenidate and amphetamine/dextroamphetamine salts (Chai et al., 2012). However, after peaking in 2004, atomoxetine use decreased from 10 million prescriptions annually to approximately 2 million per year by 2010 (Chai et al., 2012). The reasons for this decline are not entirely clear, but perceptions of variable effectiveness and the addition of an FDA boxed warning to atomoxetine labeling regarding suicidal thinking in children and adolescents in November of 2005 may be contributors. Collectively, this has resulted in the need to explore precision medicine approaches to optimize the benefits and minimize the risks while taking atomoxetine.
Atomoxetine concentrations vary significantly across patients, primarily due to genetic variation in cytochrome P450 2D6 (CYP2D6) activity (Gibson et al., 2006; Hazell et al., 2009). CYP2D6 is a well described drug metabolizing enzyme with the gene having close to 200 allele variants, which are typically reported as star (*) alleles with an assigned numeric activity value of either 0, 0.25, 0.5, or 1 associated with them, resulting in a CYP2D6 activity score (Chiba et al., 2012; Gaedigk et al., 2018). The primary metabolite of atomoxetine, 4-OH-atomoxetine, is equipotent to atomoxetine, though it circulates in comparatively low concentrations and is rapidly glucuronidated and excreted in the urine (Dinh et al., 2016). Atomoxetine exposure is 10-fold higher in CYP2D6 poor metabolizers, and maximum plasma concentrations at steady state are approximately 5-fold higher in CYP2D6 poor metabolizers when compared to normal metabolizers (Sauer et al., 2003). Additionally, the half-life of atomoxetine in poor metabolizers is considerably longer at 21.6 hours versus 5.2 hours in normal metabolizers. Large, population-level studies suggest that patients who are CYP2D6 poor metabolizers may be more likely to respond favorably to atomoxetine therapy (Michelson et al., 2007).
Given the influence of CYP2D6 on atomoxetine metabolism and exposure, genotype-guided dosing recommendations are available within the package insert (Strattera, 2002). and from the Clinical Pharmacogenetics Implementation Consortium (CPIC) guidelines when CYP2D6 genotypes are available (Brown et al., 2019). The primary objective of this medical records review was to assess the relationship between CYP2D6 phenotype and discontinuation patterns in children and adolescents with ADHD treated with atomoxetine.
Patients and Methods
A retrospective electronic health record (EHR) review was conducted for all patients 21 years or younger with a diagnosis of ADHD who received a prescription for atomoxetine from a Children’s Minnesota prescriber between January 1, 2000, and September 1, 2024, and a known CYP2D6 genotype available by September 1, 2024. Participants were excluded if the legal guardian (if participant is under 18 years of age) or the individual (if participant is 18 years or older) opted out of the general consent for research participation or if they did not have a CYP2D6 genotype. Pharmacogenomic (PGx) testing is routinely ordered at Children’s Minnesota through the Clinical Pharmacogenomics service. While testing may have been ordered to help guide atomoxetine dosing, for some patients it may have been ordered for other reasons. Although the primary reason for ordering PGx testing was not available, atomoxetine starting doses were compared in patients where CYP2D6 genotype was available before versus after the start of treatment. CYP2D6 variants tested and methodology were dependent on the individual assay. All assays included used single nucleotide polymorphism-based genotyping methods and determined the most likely diplotype based on published allele frequency. All assays, except one, included the Association for Molecular Pathology minimum set of alleles, Tier 1 and many of Tier 2.
Patient variables collected included demographics (e.g., sex, birth date, and race), atomoxetine dose and treatment duration, medication changes and discontinuation, and toxicity and efficacy markers. The initial dose was the first dose prescribed and the closest weight at the time of that prescription to determine the weight-based dosing. Any titration in dose, either up or down, was recorded separately as a dose change. The final dose was calculated based on the last prescribed dose and closest weight at the time of that prescription. Concomitant medications were also reviewed for drug-drug and drug-drug-gene interactions. CYP2D6 activity scores and phenotypes were adjusted for concomitant medications known to be moderate or strong CYP2D6 inhibitors (Cicali et al., 2021). Side effects of interest associated with atomoxetine were those reported at an occurrence of greater than 1% as documented in the package insert (Strattera, 2002). Data collected from patient EHRs was stratified according to CYP2D6 genotype-based metabolizer groups to determine associations between genotype and treatment outcome and toxicity. Patient records were reviewed for the duration of time patients were prescribed atomoxetine. Patients were followed from the index date (first drug dose) until the earliest of atomoxetine discontinuation, loss to follow-up (e.g., transfer of care), or the end of the data collection period (November 22, 2024). The primary outcome was the reason for atomoxetine discontinuation. Discontinuation events were classified into two competing risks: discontinuation due to adverse effects (i.e., intolerance/toxicity) or therapeutic failure (i.e., lack of efficacy). Patients who discontinued for reasons unrelated to these outcomes (e.g., noncompliance and treatment completion following symptom resolution) or those who were lost to follow-up were censored at the time of their last documented clinical encounter. Initial doses, dose changes at interim visits, medication changes, and terminal doses were also documented, as well as patient weight at the time of initiation and discontinuation of atomoxetine.
Patients’ CYP2D6 activity scores were adjusted to account for phenoconversion through the utilization of the University of Florida CYP2D6 phenoconversion calculator (Cicali et al., 2021) by multiplying the activity score by a conversion factor based on the patient’s original activity score and strength of the inhibitor. Phenoconversion is when a patient’s genotype-predicted metabolizer status does not match their real-world metabolizer status and may be due to a variety of factors such as concomitant medication use, smoking, and comorbidities. For example, patients taking a strong CYP2D6 inhibitor, such as paroxetine, had their CYP2D6 activity score multiplied by a conversion factor of 0 (e.g., a CYP2D6 normal metabolizer patient with a CYP2D6 activity score of 2 who was also taking paroxetine would be converted to a poor metabolizer or CYP2D6 activity score of 0 by multiplying 2 by 0). Moderate inhibitors (e.g., duloxetine) were given an inhibitor conversion factor of 0.5. No adjustment was made for weak inhibitors. Due to the small sample size and similarities in outcomes, analyses were performed with patients grouped into poor/intermediate and normal/ultrarapid groups, as well as using the adjusted activity score as a continuous predictor.
For the purpose of analysis, groupings for primary analyses were made based on adjusted activity scores. The Wilcoxon rank sum test, Pearson chi-squared test, Kruskal–Wallis test, Dunn’s test, and Fisher’s exact tests were used to assess associations with metabolizer groups. To evaluate the time to atomoxetine discontinuation, we employed a competing risks framework to account for the mutually exclusive events of discontinuation due to lack of efficacy versus discontinuation due to toxicity. Specifically, we utilized the Fine–Gray subdistribution hazard model. In this model, individuals who experienced a competing event remained in the risk set, with their weights updated to reflect the reduced risk. Regression models were adjusted to account for confounding by age, sex, race (grouped into White/Caucasian, Black/African American, and other), and year of treatment initiation. Proportional subdistribution hazards assumptions were checked using Schoenfeld residual tests and visual inspection of log-log survival curves. Results are presented as subdistribution hazard ratios (sHRs) with 95% confidence intervals. Marginal cumulative incidence curves were computed using direct standardization over the confounders with 95% confidence intervals computed through bootstrapping. Monotone penalized splines were used to model the continuous nonlinear effect of adjusted score on hazard to discontinue. The significance level was set at 5%, and when multiple comparisons were made, p-values were adjusted using the Holm procedure. Analyses were performed in R.
Results
One hundred and eight patients, 31 females (29%) and 77 males (71%), met inclusion criteria, 86 (80%) of whom identified as White/Caucasian, 17 (16%) as Black/African American, 2 (1.9%) as Native Hawaiian/Pacific Islander, and 6 (5.6%) as Hispanic/Latino, with a median age of 9.3 years (7.5–12.3 interquartile range [IQR]). All patients were diagnosed with ADHD, with several also having a diagnosis of autism (N = 7, 6.5%), anxiety (N = 1; 0.9%), adjustment disorder (N = 1; 0.9%), mood disorder (N = 1; 0.9%), and/or oppositional defiant disorder (N = 1; 0.9%). Based on CYP2D6 genotype prior to adjustment of activity score for concomitant medications, there were 5 (4.6%) poor metabolizers, 33 (31%) intermediate metabolizers, 66 (61%) normal metabolizers, and 4 (3.7%) ultrarapid metabolizers. After adjusting activity scores and CYP2D6 phenotype to account for phenoconversion, there were 13 (12%) poor metabolizers, 30 (28%) intermediate metabolizers, 61 (56%) normal metabolizers, and 4 (3.7%) ultrarapid metabolizers, totaling 43 (40%) patients in the poor/intermediate metabolizer group and 65 (60%) in the normal/ultrarapid group (Table 1 and Supplementary Table S1). Concomitant use of moderate or strong CYP2D6 inhibitors was identified in 10 patients (9.2%) within the study cohort, all of whom had their CYP2D6 activity score adjusted. The most frequently prescribed strong inhibitor was fluoxetine (N = 7, 6.5%). Other medications included the moderate inhibitor duloxetine (N = 2, 1.9%) and the strong inhibitor bupropion (N = 1, 0.9%). The two metabolizer groups differed significantly only in cardiovascular and neurological adverse effects, with the poor/intermediate metabolizers experiencing significantly more adverse effects than the normal/ultrarapid group (Table 1 and Supplementary Table S1). See Supplementary Table S2 for the full list of adverse effects which were surveyed.
Demographics and Clinical Characteristics by Adjusted Activity Score Group
Median follow-up time for the cohort, calculated using the reverse Kaplan–Meier method, was 5.53 years (95% CI: 2.8–8.82). The total follow-up accrued was 208 person-years, during which 78 patients (72%) of the cohort reached the primary endpoint of treatment discontinuation.
Compared to the poor/intermediate group, patients in the normal/ultrarapid group were significantly more likely to discontinue due to lack of efficacy (sHR = 1.9, p = 0.04). After one year following initiation of therapy, the estimated cumulative incidence of discontinuation for lack of efficacy was 29.5% (95% CI: 20.2–41.1) in the normal/ultrarapid group compared to 16.9% in the poor/intermediate group (Fig. 1A). Conversely, patients in the normal/ultrarapid group were significantly less likely to discontinue due to toxicity (sHR = 0.4, p = 0.03). The 1-year cumulative incidence of discontinuation due to toxicity was 19.1% (95% CI: 8.8–29.6) for the normal/ultrarapid group compared to 36.9% (95% CI: 23.0–51.7) for the poor/intermediate group (Fig. 1B).
Cumulative incidence functions for atomoxetine discontinuation by adjusted activity score group. Facets show the probability to discontinuation from
Comparisons were also made using the activity score directly as a continuous predictor of time to discontinue, which showed a significant nonlinear effect for both lack of efficacy (p < 0.001) and for toxicity (p < 0.001). This approach shows an increasing probability to discontinue due to the lack of effect with higher activity scores and a sharper rise with activity scores greater than 2 (Fig. 2). The model shows a decreasing probability of discontinuing due to toxicity as the activity score increases, starting high with a sharp decrease, followed by a plateau between scores 1 and 2 and then a more gradual decrease for scores greater than 2.
Nonlinear relationship between adjusted activity score and the probability of discontinuation after one year from start of therapy due to
A Kruskal–Wallis test demonstrated a significant difference in initial weight-based doses across the four cohorts defined by activity group and genotype availability at the time of prescription (p < 0.001). Post hoc analysis using Dunn’s test revealed that patients whose genotype was available prior to prescribing and who were identified as intermediate or poor metabolizers received a significantly lower median dose (0.5 mg/kg/day; IQR 0.4–0.6) than those whose status was known to be normal or rapid (0.9 mg/kg/day; IQR 0.6–1.1; p = 0.003). This cohort also received significantly lower initial doses compared to both groups where the genotype was obtained after the initial dose, specifically the unknown intermediate or poor metabolizer group (0.7 mg/kg/day; IQR 0.5–1.0; p = 0.025) and the unknown normal or rapid metabolizer group (1.0 mg/kg/day; IQR 0.6–1.2; p < 0.001). Conversely, there were no significant differences in dosing between the group with a known normal or rapid status and either of the unknown genotype cohorts (p > 0.05 for all comparisons). Furthermore, in the absence of pre-prescription genotyping, initial doses did not differ significantly between patients who were eventually identified as intermediate/poor versus normal/rapid metabolizers (p = 0.42). These findings suggest that dosing was more conservative when a reduced-function genotype was identified before the initial prescription (Fig. 3).
Comparison of initial atomoxetine dose (mg/kg/day) by adjusted activity score group and genotype availability. Initial normalized dose grouped by initial adjusted score group and
A statistically significant difference was observed between the adjusted phenotype group and their last recorded atomoxetine doses (p = 0.01). The poor/intermediate phenotype group had a median final dose of 0.8 mg/kg/day (IQR 0.4–1.2), while the normal/ultrarapid group had a median final dose of 1.0 mg/kg/day (IQR 0.7–1.3; Fig. 4A). Similarly, final atomoxetine doses differed significantly based on discontinuation status (p = 0.001). Post hoc pairwise comparisons revealed that patients who discontinued due to toxicity (N = 31; median = 0.7 mg/kg/day) received significantly lower final doses than both those who discontinued due to lack of efficacy (N = 47; median = 1.0 mg/kg/day; p = 0.005) and those who were censored (N = 30; median = 1.2 mg/kg/day; p = 0.002). No significant difference in final dose was observed between patients who discontinued due to lack of efficacy and those who were censored (p = 0.32; Fig. 4B).
Association between final atomoxetine dose (mg/kg/day) and adjusted activity score group and discontinue status
Discussion
Although not a first-line medication for ADHD in children, atomoxetine remains an important second-line treatment for parents/children looking for non-stimulant options or those who are unable to tolerate adverse effects experienced with stimulants. CYP2D6 poor metabolizers have previously been shown to respond better to atomoxetine as compared to individuals who are not poor metabolizers in some (Michelson et al., 2007), but not all (Bishop et al., 2024) studies. In our retrospective review, children with ADHD taking atomoxetine and who had undergone PGx testing were shown to have significantly different discontinuation patterns depending on their CYP2D6 phenotype. Children who were poor or intermediate metabolizers of CYP2D6 were significantly more likely to discontinue atomoxetine due to toxicity as compared with normal or ultrarapid metabolizers, while children who were normal or ultrarapid metabolizers were significantly more likely to discontinue due to lack of efficacy. Additionally, as adjusted CYP2D6 activity score increased, a trend toward discontinuation of atomoxetine due to lack of efficacy was observed. Conversely, as adjusted CYP2D6 activity score decreased, a trend toward discontinuation of atomoxetine due to toxicity was also observed. These data collectively support previous observations suggesting that CYP2D6 phenotype is associated with treatment outcomes and that PGx testing for CYP2D6 should be used to guide atomoxetine dosing.
A recent retrospective review of 315 children, adolescents, and adults taking atomoxetine with serum concentrations demonstrated that CYP2D6 intermediate metabolizers experienced nearly a two-fold higher exposure when compared to CYP2D6 normal metabolizers, while poor metabolizers experienced nearly 10-fold greater exposure (Smith et al., 2023). Additionally, a large study in Chinese children (N = 385) offers further support that individuals with reduced CYP2D6 activity have greater atomoxetine exposure and are more likely to respond. Guo et al. reviewed PGx and therapeutic drug monitoring results of children with ADHD taking atomoxetine, showing that intermediate metabolizers had higher exposure and greater response as compared to normal metabolizers (Guo et al., 2024). Conversely, a double-blind crossover study in the United States showed that children who are CYP2D6 intermediate metabolizers have less symptom improvement at 3 weeks compared to normal metabolizers but no difference after 4 weeks (Bishop et al., 2024).
Notably, prescriber behavior seemed to be influenced by whether or not PGx information was available prior to the start of atomoxetine. No differences in starting dose were noted when CYP2D6 genotype was not available prior to starting atomoxetine between intermediate/poor and normal/ultrarapid metabolizers; however, when CYP2D6 genotype was available, prescribers were more conservative in initial dosing with intermediate/poor metabolizers, likely associated with clinical decision support at the time of prescribing that aligns with the CPIC recommended dose of 0.5 mg/kg/day, as compared to normal/ultrarapid metabolizers. While PGx testing was likely not specifically ordered to inform atomoxetine dosing, this institution has had a clinical PGx service for nearly a decade, so prescribers may have greater familiarity and comfort levels in how to apply PGx-guided information when available (Paetznick et al., 2022).
There are a number of limitations that need to be considered when interpreting the results of this study. This study is limited by its retrospective nature and predominantly White/Caucasian population from a single site, making it difficult to infer causality between genetic findings and treatment outcomes. In rare cases, children may have been on atomoxetine upon entering the health system and thus were not a true new atomoxetine start. We were also unable to include prior medication exposure, which should be considered in future studies, and CYP2D6 poor/intermediate metabolizers were grouped and compared with normal/ultrarapid metabolizers based on the limited available sample size. Therapeutic drug monitoring was also not conducted on any of the patients included in our analyses. This limited comparisons between efficacy and tolerability to atomoxetine doses and not necessarily resulting exposures. Additional PGx results that may contribute to atomoxetine response were also not available, namely the norepinephrine transporter gene (SLC6A2). Variants in SLC6A2 have been shown to be associated with atomoxetine response, suggesting that future research should focus on both exposure and genetics (Ramoz et al., 2009).
Conclusion
These data suggest that CYP2D6 phenotype may be associated with discontinuation patterns in children with ADHD treated with atomoxetine. Additionally, the availability of PGx results also appears to impact how prescribers select atomoxetine dosing. Clinicians should consider ordering PGx testing when a child’s CYP2D6 genotype is not available prior to prescribing atomoxetine in order to help guide dosing and interpret clinical response. Knowing CYP2D6 genotypes prior to starting atomoxetine could help to optimize dosing strategies while preventing unnecessary treatment discontinuation.
Authors’ Contributions
J.T.B., E.N.D., J.R.B., and D.B.G.: Designed the research; E.N.D., and L.Y.: Curated the data; B.D., and D.A.W.: Performed statistical analysis; J.T.B., E.N.D., C.P., J.R.B., B.D., D.A.W., and D.B.G.: Analyzed the data.
Supplemental Material
sj-docx-1-cha-10.1177_10445463261467214 — Supplemental material for Association of Atomoxetine Discontinuation Patterns with CYP2D6 Phenotype in Children and Adolescents with Attention Deficit Hyperactivity Disorder
Supplemental material, sj-docx-1-cha-10.1177_10445463261467214 for Association of Atomoxetine Discontinuation Patterns with CYP2D6 Phenotype in Children and Adolescents with Attention Deficit Hyperactivity Disorder by Jacob T. Brown, Elise N. Durgin, Courtney Paetznick, Jeffrey R. Bishop, Benjamin Deonovic, Laylia Yang, David A. Watson, and David B. Gregornik
Supplemental Material
sj-docx-2-cha-10.1177_10445463261467214 — Supplemental material for Association of Atomoxetine Discontinuation Patterns with CYP2D6 Phenotype in Children and Adolescents with Attention Deficit Hyperactivity Disorder
Supplemental material, sj-docx-2-cha-10.1177_10445463261467214 for Association of Atomoxetine Discontinuation Patterns with CYP2D6 Phenotype in Children and Adolescents with Attention Deficit Hyperactivity Disorder by Jacob T. Brown, Elise N. Durgin, Courtney Paetznick, Jeffrey R. Bishop, Benjamin Deonovic, Laylia Yang, David A. Watson, and David B. Gregornik
Footnotes
Author Disclosure Statement
The authors have no financial conflicts of interest to disclose.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
