As shown in Table 3, the inter-individual variability in the plasma clearance of atomoxetine, corrected by oral bioavailability (CL/F), ranges between 14%–62% CV in children. This variability is closely associated with the CYP2D6 phenotype. Significant differences in plasma atomoxetine concentrations among children exist, depending on the route of administration (Guo et al., 2024). Most recently, Guo et al. (2024) identified sex, BW, and CYP2D6 phenotype were the primary factors influencing individual exposure to atomoxetine, with the phenotype exerting the most significant impact. Although the impact of CYP2D6 phenotype on pharmacokinetics of atomoxetine in children has been observed, studies in this area remains quite limited.

Currently, reports on the intra-individual differences in pharmacokinetic parameters, such as plasma atomoxetine concentrations and total CL, are very limited. Recently, Cheng et al. (2023) utilized a population pharmacokinetic (PPK) modeling approach to estimate the residual unexplained variability (i.e., intra-individual variability) in plasma concentrations of atomoxetine and its major metabolite, 4-OH-atomoxetine, in children and adolescents, yielding 21.3% and 29.6% CV, respectively. In addition, early studies have indicated that, food intake reduces its peak concentration and delays the time to peak concentration, although it does not affect the absorption of atomoxetine (Sauer et al., 2005). This factor should also be considered as contributing to both intra-individual and inter-individual variability.

When ADHD occurs alongside other conditions such as anxiety or depression, there may be a need for using atomoxetine in combination with other medications; however, this is more common in adults (Todor et al., 2017; Todor et al., 2016; Belle et al., 2002; Sauer et al., 2004; Kratochvil et al., 2005). Research has evaluated the impact of bupropion, fluvoxamine, paroxetine, desipramine, and fluoxetine on the pharmacokinetics of atomoxetine, as a victim drug (Table 3). Studies involving fluvoxamine (Todor et al., 2017) and desipramine (Sauer et al., 2004) indicated that interactions, if present, were slight; however, fluoxetine increased atomoxetine peak concentration by 3.4 fold (Kratochvil et al., 2005). Of note, bupropion exhibited significant inhibitory effects on atomoxetine’s metabolism in patients classified as CYP2D6 EMs, while the effects were minimal in CYP2D6 PMs (Todor et al., 2016). Interestingly, a Canadian guideline classifies bupropion as a third-line treatment for ADHD (Schoretsanitis et al., 2019). In addition, a case report has noted an improved response to atomoxetine in a patient, likely classified as a CYP2D6 EM, following the addition of paroxetine (Paulzen et al., 2016), indicating that these inhibitors can be utilized to enhance the response to atomoxetine in individuals identified as CYP2D6 EMs.

Conversely, a recent study involving children and adolescent with ADHD revealed that the use of concomitant medications is quite rare, particularly regarding herbal medicines (Guo et al., 2024). Nevertheless, instances in adults where the concurrent use of medications as potential inhibitors of CYP2D6 (Table 4) significantly alters the systemic exposure to atomoxetine warrant careful consideration by pediatricians. If similar combinations of medications become necessary for children and adolescents, it may be important to adjust doses to account for changes in exposure due to DDIs (Sauer et al., 2004) in order to ensure efficacy and minimize adverse reactions (Fu et al., 2023).

It is crucial to highlight that whether CYP2D6 inhibitors have a practical effect is closely linked to the patient’s CYP2D6 metabolic phenotype. They do not work in CYP2D6 PMs but can enhance systemic exposure to atomoxetine in CYP2D6 non-PMs. Therefore, gathering genotype and phenotype information about the patient’s CYP2D6 status becomes necessary.

As of now, the therapeutic window for atomoxetine in treating ADHD in children has not been clearly established. The recommended concentration range of 200–1,000 ng/mL, as outlined in guidelines (Hiemke et al., 2018; Brown et al., 2019), primarily focuses on identifying the lowest concentration necessary to achieve efficacy. However, the link between plasma levels of atomoxetine and its clinical effectiveness remains unclear, complicating the relationship between tolerability and concentration even further (Guo et al., 2024). Consequently, it is more plausible that we will first define a minimum concentration required for therapeutic effect, while finding a corresponding concentration that ensures tolerability is challenging, as tolerability does not always have a straightforward relationship with drug concentration.

Atomoxetine is taken orally, and main pharmacokinetic parameters identified in the literature include plasma peak concentration (C _max ), AUC, and CL/F (Table 3), as along with plasma concentrations measured at specific intervals after administration (e.g., 12 h). Current clinical evidence strongly associates C _max with the efficacy of atomoxetine, which is why it is recommended as a primary monitoring parameter in guidelines (Hiemke et al., 2018; Brown et al., 2019). Although establishing a connection between C _max and adverse reactions can be more difficult, some findings in the literature address this relationship as well (Guo et al., 2024).

A systematic review and dose-response meta-analysis found that the effectiveness of atomoxetine increased up to a dosage of 1.4 mg/kg, after which it plateaued (Terao et al., 2024). There is considerable interest in determining if specific plasma concentrations of atomoxetine can predict the level of clinical response. In an early investigation, Michelson and co-researchers applied a nonlinear model to analyze peak concentrations and the relative change from baseline in the ADHDRS-IV-Parent: Inv total score. This model indicated that the maximum expected improvement compared to baseline would be −23.5, aligning with a plasma atomoxetine concentration of 400 ng/mL (Michelson et al., 2007). However, Hazell et al. revealed that while certain patients may benefit from higher plasma atomoxetine levels (>800 ng/mL), mere exposure to these levels dose not reliably predict the therapeutic outcomes in children with ADHD, suggesting that other factors also influence the response to atomoxetine (Hazell et al., 2009).

In a recent non-randomized prospective interventional study, Sugimoto et al. (2021) found that children with ADHD aged 6–12 years (n = 43) were more likely to respond to respond to atomoxetine treatment when its steady-state plasma concentration exceeded 64.60 ng/mL. Similarly, Guo et al. (2024) identified a lower threshold of 268 ng/mL as a potential therapeutic reference range for pediatric patients receiving q.m. atomoxetine, suggesting that effectiveness increases when this level is surpassed. Conversely, Ruppert et al. (2022) found that neither a concentration-effect relationship nor a dose-effect relationship was observed.

In general, CYP2D6 PMs are more likely to experience side effects from atomoxetine than non-PMs, likely due to their higher exposure to the drug (Brown et al., 2019; Michelson et al., 2007). However, it is still uncertain whether drug exposure metrics like C _max or AUC have a significant influence on tolerability, as there are only a few studies investigating the relationship between plasma atomoxetine concentrations and clinical outcomes. One early clinical trial found no correlation between plasma atomoxetine concentrations and its tolerability (Hazell et al., 2009). Similarly, a TDM study involving children and adolescents with ADHD did not reveal any clear relationship between serum concentrations and side effects (Ruppert et al., 2022). In contrast, a recent study by Guo et al. (2024) did identify a correlation between certain adverse reactions and plasma atomoxetine concentration. Specifically, in CYP2D6 IMs receiving once-daily dosing or EMs receiving twice-daily dosing, a significant difference was observed in the occurrence of gastrointestinal (e.g., 510 vs. 386 ng/mL, p = 0.0411) and neurological adverse reactions, even at plasma atomoxetine concentrations where no adverse reactions were reported.

As of now, no studies have directly compared the therapeutic effects of atomoxetine before and after the implementation of TDM. However, clinicians at our hospital believe that TDM has significantly enhanced their ability to select medications and adjust dosages. With TDM support, they can make more timely clinical decisions, such as switching from atomoxetine to alternative medications or tailoring the dosage. This approach has allowed for a more efficient determination of the optimal dosage for pediatric patients using atomoxetine. Notably, some children with ADHD have experienced effective control with relatively lower doses, an outcome that was less common prior to TDM implementation. We are currently gathering such real-world clinical data and planning to design clinical trials to systematically evaluate the benefits of implementing TDM for atomoxetine.

Toxicity from overdose is thought to arise from elevated synaptic NE levels, which can induce an excessive noradrenergic-mediated sympathomimetic syndrome, which typically presents as tachycardia and hypertension. In cases of atomoxetine overdose, clinical manifestations are generally mild. Common symptoms include drowsiness (particularly in children), agitation, hyperactivity, gastrointestinal disturbances, tremors, hyperreflexia, tachycardia, hypertension, and seizures. Fortunately, these symptoms typically resolve quickly, with complete recovery usually occurring within 24 h post-overdose (Spiller et al., 2013). However, some patients may need to discontinue atomoxetine due to inability to tolerate several common adverse reactions reported in clinical trials, including nausea, vomiting, fatigue, decreased appetite, abdominal pain, and somnolence (Mechler et al., 2022).

Similarly, there has been no public report to date examining whether the tolerability of atomoxetine improves before and after the implementation of TDM. Additionally, the correlation between tolerability and concentration has also not been established. However, it is indeed more likely for PMs to experience adverse reactions compared to non-PMs (Brown et al., 2019). From this perspective, implementing TDM is expected to improve tolerability. For example, TDM could help identify children on low doses of atomoxetine who have low exposure, resulting in poor tolerance and inadequate efficacy. In such cases, a timely medication switch may be appropriate, potentially avoiding the need for further dose escalation to achieve efficacy. Conversely, for children with high exposure who show good efficacy and good tolerance, a dose reduction can be considered to alleviate the body’s burden of atomoxetine. This area warrants exploration in clinical trial and represents a significant clinical issue that should be prioritized.

Various bioanalytical assays have been established to measure atomoxetine in human plasma, serum, urine, or hair, using techniques such as (high-performance) liquid chromatography combined with detection methods such as UV detector (Patel et al., 2007; Guo et al., 2007; Teichert et al., 2020), fluorescence detector (Stegmann et al., 2016; Zhu et al., 2007), or by (tandem) mass spectrometry (Mullen et al., 2005; Papaseit et al., 2012; Papaseit et al., 2013; Sim et al., 2017; Choi et al., 2012; Marchei et al., 2012; Xia et al., 2021; Skaalvik et al., 2021). Most of these methods are lab-developed and may have limited general applicability. Recently, new strategies have emerged recently. For example, Abu-Hassan developed a Nano-level assay based on molecular-size-based resonance Rayleigh scattering to detect atomoxetine in both its prescribed dosage form and plasma samples. This environmentally friendly fluorometric technique shows considerable promise for application due to its significant advantages, such as intelligent selectivity, exceptional sensitivity, minimal solvent consumption, widespread availability in laboratories, rapid analysis times, and ease of use (Abu-Hassan, 2023). Importantly, the choice of method is less critical than ensuring accurate determination of atomoxetine concentration in biological samples; researchers can select an assay based on its accessibility.

When performing TDM for atomoxetine, several factors need to be considered regarding blood collection methods. First, the genetic polymorphism of CYP2D6 and metabolic phenotypes (UM, NM, IM, and PM) of CYP2D6 significantly influence the drug’s metabolism, leading to significant differences in its t _1/2. Notably, the t _1/2 of atomoxetine in PMs was 4-fold higher than that of EMs (Sauer et al., 2005; Michelson et al., 2007; Byeon et al., 2015). In clinical practice, C _max is primarily used as a parameter to assess the correlation between systemic exposure to atomoxetine, its effectiveness, and adverse reactions. Consequently, patients with different CYP2D6 phenotypes may experience varying peak times even under the same dosing regimen. Given the pharmacokinetic variations linked to CYP2D6 phenotypes that affect C _max and t _1/2, the CPIC guideline recommends that prescribers consider measuring peak concentrations at specific time intervals: 1) 1–2 h post-dose in known CYP2D6 UMs, NMs, and IMs with high activity (AS 1.0 without the CYP2D6*10 allele); 2) 2–4 h post-dose in CYP2D6 IMs with low activity (AS 0.5) and individuals with an AS of 1 who carry the CYP2D6*10 allele; and 3) 4 h post-dose for PMs (Brown et al., 2019).

Second, the dosing regimen also affects the timing blood sample collection. For q.m. and b.i.d. regimens, it's generally straightforward to collect blood samples 1–4 h after drug administration. However, for children who take medication once at night (q.n.; not many, but seen (Mechler et al., 2022)), determining the interval (e.g., 12 h) (Guo et al., 2024; Sugimoto et al., 2021) for blood sample collection can be a challenging issue. This situation may necessitate prior communication with the physician regarding the timing of the previous night’s medication and the blood collection time the following day.

Third, if genetic and phenotypic information is not available, or even if it is, a concentration obtained from a single time-point sample may not accurately represent the C _max. The CPIC guideline also suggest collecting blood sample within a specific time window after dosing (Brown et al., 2019). Therefore, it may be necessary to consider sampling at steady state, despite studies that investigate the relationship between steady-state trough concentrations and clinical response (Sugimoto et al., 2021).

As of now, two guidelines provide recommendations for the therapeutic reference range of atomoxetine. According to the AGNP TDM Expert Group consensus guidelines, peak plasma concentrations between 200 and 1,000 ng/mL, measured 60–90 min after a dose of 1.2 mg/kg/day, are commonly regarded as the therapeutic reference range, but this has only been studied in adults (Hiemke et al., 2018). The latest CPIC guideline also establishes a therapeutic reference range for peak plasma concentration at 200 and 1,000 ng/mL, noting that adequate responses can be achieved when the plasma concentrations exceed 400 ng/mL (Brown et al., 2019).

Interestingly, a recent retrospective study by Guo et al. involving children with ADHD, recommended a minimum C _max of 268 ng/mL associated with achieving a favorable therapeutic effect for patients receiving q.m. dosing of atomoxetine (Guo et al., 2024). Additionally, a naturalistic study in children and adolescents with ADHD proposed a therapeutic reference range of 100–400 ng/mL (Sugimoto et al., 2021). Researchers also recommend the minimum steady-state trough concentration of 64.6 mg/mL necessary for a good control of ADHD symptoms (Sugimoto et al., 2021).

In 2019, the CPIC released guidelines proposing the use of plasma concentration in conjunction with an individual’s CYP2D6 genotype to assist clinicians in dose selection and titration. For patients classified as CYP2D6 UMs and NMs, if the peak concentration is <200 ng/mL and there is no clinical response, it is advisable to increase the dose proportionately to achieve approximately 400 ng/mL. For CYP2D6 PMs, IMs, and NMs with an AS of 1 who carry the CYP2D6*10 allele and taking a standard starting dose, the recommendation is to consider a proportional dose adjustment to reach about 400 ng/mL if there is an inadequate response without side effects (Brown et al., 2019).

In recent years, there have been extensive efforts to create personalized dosing strategies for atomoxetine using PBPK (Shimizu et al., 2023; Rudesheim et al., 2022; Dinh et al., 2016; Kim et al., 2018; Notsu et al., 2020; Alsmadi et al., 2022) and PPK (Cheng et al., 2023) models. Of note, recent PPK simulations revealed that the majority of individuals with a CYP2D6 AS of 1–3 may not achieve a steady-state C _max of 400 ng/mL with a 0.5 mg/kg once daily (q.d.) dosage, whereas most individuals with a CYP2D6 AS<1 could reach this concentration. This suggests that individuals with CYP2D6 AS of 1–3 may require a higher dose of atomoxetine compared to those with scores <1. To achieve a steady-state atomoxetine C _max comparable to that of individuals with an AS of 0 following a 0.5 mg/kg q.d. dose of atomoxetine, individuals with CYP2D6 AS 1-3 would require an approximately 1.2 mg/kg q.d. dose (Cheng et al., 2023). These findings largely align with the dosing recommendations outlined in the above noted CPIC guideline (Brown et al., 2019).

Cost-effectiveness analysis in healthcare, particularly for TDM, is still developing. Initially, TDM was only shown to be cost-effective for aminoglycosides (Touw et al., 2005). However, recent evidence indicates that TDM interventions can also be cost-effective in the application of antibody drugs (Martelli et al., 2017) and anti-cancer medications (Vithanachchi et al., 2021). While there is some rationale supporting the TDM of atomoxetine, comprehensive cost-effectiveness analyses have not yet been conducted. Consequently, the emphasis should extend beyond just cost-effectiveness to encompass how these interventions can be implemented in a clinically beneficial and economically sustainable way.

When integrating pharmacogenomics (PGx) into clinical practice, cost is also a crucial consideration for both healthcare systems and patients (Morris et al., 2022). Despite a substantial decrease in of PGx testing costs over the past decade, it continues to pose a significant barrier to widespread implementation in children’s hospitals (Brown et al., 2021). For certain medications, such as clopidogrel and warfarin, there is considerable cost data available that provide strong support for the use of PGx testing (Dong et al., 2020; Zhu et al., 2021). Additionally, cost-effectiveness analyses of CYP2D6 genotyping have primarily focused on antidepressant medications (Groessl et al., 2018; Maciel et al., 2018), with no relevant studies on atomoxetine thus far.

For certain medications, integrating TDM with pharmacogenomics proves to be an effective approach to optimize treatment, emphasizing the importance of assessing the cost-effectiveness of both methods. One study found that a combined strategy of NUDT15/TPMT genotype screening prior to initiating azathioprine treatment, along with on-going TDM for management, was more cost-effective than alternatives that involved either genotyping NUDT15 or TPMT alone or conducting genotyping without TDM in patients with inflammatory bowel disease (Zeng et al., 2021). However, comprehensive cost evaluations for atomoxetine are still lacking.

The first challenge in implementing personalized dose adjustment lies in the unclear relationship between atomoxetine exposure levels and both its efficacy and adverse reactions. ADHD is a complex and heterogeneous disorder (Posner et al., 2020; LaBianca et al., 2024), highlighting the need to evaluate medication responses in relation to the pharmacokinetics and duration of action of the selected formulation. Optimal symptom management and functional improvement occur when blood levels of the medication are adequately maintained for the periods of greatest need and for the specific tasks at hand (Faraone et al., 2024). Current evidence generally supports identifying the lowest concentration or concentration range that ensures optimal efficacy; however, data regarding the maximum tolerable concentration levels still relatively sparse. Additionally, there is also scarce data on the relationship between exposure levels to atomoxetine and its adverse reactions. Consequently, there is no clear therapeutic window defined for atomoxetine. Without this defined window, there is no established “target value” for dose selection, complicating the process of making personalized dose adjustments. Also, the lack of a defined therapeutic window has compromised the role of PPK/PBPK models in predicting personalized doses, posing a significant challenge for future efforts in this area. In response to these challenges, machine learning-based predictive models have emerged as a promising strategy (Faraone et al., 2021; Faraone et al., 2022).

The second challenge involves the lack of predictors for treatment response to atomoxetine. By integrating biomarkers and clinical predictors of both response and adverse effects, clinicians could potentially tailor treatment for individual patients. However, there are currently no available clinical or biological predictors of response for ADHD (Buitelaar et al., 2022). At this time, the genotypes and phenotypes of CYP2D6 may serve as the most “reliable” predictor. As the primary metabolic enzyme for atomoxetine, CYP2D6’s metabolic activity directly influences the drug’s pharmacokinetic behavior, thereby linking exposure levels to both efficacy and adverse reactions. In other words, variations in CYP2D6 activity fundamentally “determine” the differences in both the efficacy and adverse reactions of atomoxetine. However, the predictive power of CYP2D6 is limited, mainly due to the lack of a well-established exposure-response relationship, as previously noted. Additionally, the inherent uncertainties in predicting drug response based solely on genetic markers, given the potential for false negatives or positives in CYP2D6 genotyping, also pose challenges in clinical practice as well. Investigating ways to standardize the translation of genotyping data into actionable, evidence-based prescribing decisions is an important endeavor. Nonetheless, personalized dosing strategies that utilize CYP2D6 stratification remain crucial and significantly continue to the rational use of atomoxetine (Brown et al., 2019; Guo et al., 2024).

The third challenge comes from how to objectively and accurately assess the clinical efficacy of atomoxetine (Wong et al., 2019; Raman et al., 2018). To tackle this issue, we require more objective and quantifiable indicators that can accurately depict changes in symptoms and reflect treatment outcomes. Currently, it is recognized that the diagnostic rate for female ADHD patients is lower than that for males (Martin, 2024). If this discrepancy continues to affect efficacy assessments, it could hinder the effective implementation of personalized dosing strategies.

As early as 2015, it was clearly understood that achieving precision medicine would necessitate overcoming major challenges across various domains, including technological and socio-political aspects. TDM and pharmacogenomics, as key elements of precision medicine, face few technical hurdles; however, socio-political factors such as public support, affordability, and education pose even more obstacles (Kohane, 2015). Indeed, numerous challenges will persist in clinical practice, limiting the widespread implementation of precision medicine in clinical settings (Chenoweth et al., 2020).

While TDM and genotyping technologies do not inherently pose obstacles to implementing personalized medicine, their widespread use in clinical settings is not encouraging, particularly in children’s hospitals (Brown et al., 2021; Chenoweth et al., 2020; Just et al., 2019; Duarte et al., 2021), remains limited. Few institutions have the capability to conduct both TDM and genetic testing simultaneously, hindering individualized dose adjustments. For example, a nationwide survey by Jacob et al. found that only four centers implemented TDM for atomoxetine alongside CYP2D6 genotyping (Brown et al., 2021). Similarly, our recent survey on the clinical implementation of PGx testing revealed that only four children’s hospital conducted CYP2D6 genotyping (Wu et al., 2024). Additionally, the cost associated with implementing genotyping and TDM in clinical practice raises significant concern for healthcare systems and patients, making cost evaluations critical (Morris et al., 2022). Furthermore, effective implementation also requires collaboration among genotyping and TDM labs, bioinformatics/IT for result analysis and communication, and clinicians for integrating patient care. Notably, only one center offered clinical decision support for atomoxetine and CYP2D6 interaction within its electronic prescribing system (Brown et al., 2021). Therefore, building multidisciplinary teams around personalized dosing is crucial (Barker et al., 2022); where such teams are in place, they can significantly enhance the effectiveness of personalized dosing strategies.

Ultimately, a comprehensive understanding of the ADHD itself may present the greatest challenge. Our unwavering goal is to pursue personalized atomoxetine treatment, which depends on a deeper understanding of the condition’s heterogeneity through extensive studies into its etiology, pathophysiology, and clinical manifestations. Integrating multi-omics studies can facilitate the discovery and validation of biomarkers that could serve as potential clinical predictors of response (Hubers et al., 2024; Hagenbeek et al., 2023).

Additionally, it is essential to address medication adherence. Once adherence issues are resolved (Brikell et al., 2024), prioritizing the establishment of a precise match between patients and optimal atomoxetine treatment will become a focus for future research.

Overall, while this study extensively references various guidelines and theoretical frameworks, it lacks specific examples of clinical outcomes derived from these approaches. The key challenge is to effectively translate these theoretical concepts into practical, real-world clinical applications. Nonetheless, clinicians can gain valuable insights from the existing evidence, particularly in identifying the limitations of current guidelines and implementing personalized treatment across different clinical settings.