Clinical predictors of stimulant efficacy in adults with ADHD

Abstract

Background:

Attention-deficit/hyperactivity disorder (ADHD) is a disorder marked by inattentiveness and/or hyperactivity and increased impulsivity. A common treatment for ADHD is stimulant medications, with a formulation of methylphenidate or amphetamine. Although stimulant medication is effective in most patients, 40% show no response. There have been attempts to predict stimulant efficacy through neuroimaging and electroencephalograms; however, these methods are expensive and not sustainable in day-to-day clinical practices.

Objectives:

This study aimed to identify clinical factors that could be used to predict which patients would best respond to stimulants.

Design:

The reporting of this study conforms to the STROBE statement. This was a naturalistic prospective observational study.

Methods:

Thirty-six medication-naïve adults with ADHD were prescribed stimulant medication and naturalistically followed for an average of 116 days. Demographics, type of stimulant, and seven clinical rating scales were analyzed to identify response predictors. Truncated Poisson regressions, stepwise logistic regressions, receiver operating characteristic curve analysis, and Bonferroni corrections were performed.

Results:

Executive function impairment and better quality of life were found to be the best indicators for stimulant response. Higher scores on Adult Self-Report (ASR) Thought Problems, Withdrawn Problems, Internalizing Problems, and Intrusive Thoughts were indicative of lower stimulant efficacy. Poorer working memory and task monitoring also predicted lower stimulant response.

Conclusion:

These clinical measures could aid clinicians and patients in predicting who would better respond to stimulant medications and reduce the time that patients wait before finding an effective treatment. Executive functioning, quality of life, and ASR profile measurements can be used to best manage ADHD symptomology.

Introduction

Attention-deficit/hyperactivity disorder (ADHD) is a highly prevalent disorder that has significant impact at both an individual and societal level and is estimated to affect roughly 3%–6% of the adult population worldwide^1,2. The disorder can be characterized by inattentiveness and/or hyperactivity and increased impulsivity, and such impairment can lead to difficulties such as underachievement, unemployment, and antisocial behaviors. ³

Stimulant medications are most commonly used for the treatment of ADHD, with formulations including methylphenidate (MPH) or amphetamine (AMPH) found to often be safe and effective in many patients^4–7. Stimulant treatment not only improves symptoms of ADHD but can also be beneficial for educational achievements, is linked to lower antisocial behaviors, and can reduce comorbid emotional dysregulation^8–10. However, stimulant medication does not lead to improvements in all individuals with ADHD. In trials of MPH in adults with ADHD, response rate is often around 60%, indicating that 40% of individuals will not adequately respond to initial stimulant treatment^11,12.

Currently, clinicians do not have a way to predict whether their patient with ADHD will respond well to stimulant treatment or fall into the sizable percentage of non-responders. Without evidence of clinical predictors, clinicians often will use a trial-and-error method through various pharmacotherapies and stimulant formulations to determine what medication their patient will best respond to. This strategy can become protracted through efforts to try to taper medications and restabilize treatment regimens during the process.

Precision medicine has aimed to identify various predictors of treatment response to mitigate the effects of this strategy across a variety of psychiatric disorders. Ongoing research has explored the use of biomarkers and genetics in predicting response to treatment in major depressive disorder, bipolar disorder, schizophrenia, and autism spectrum disorder (ASD). For example, one study of bipolar patients identified polymorphisms in the gene Phospholipase C Gamma 1 that occurred at a significantly higher frequency in the group that responded positively to lithium treatment compared to those who were non-responders ¹³ . However, genetic polymorphisms identified as possible outcome predictors have often not yet been translatable to clinical practice ¹⁴ . Clinical predictors identified through assessment at baseline may be more feasible to collect and apply in current clinical practices. For example, in ASD, higher baseline scores on the irritability subscale of the Aberrant Behavior Checklist (ABC) were found to be associated with a larger decrease in irritability after treatment with risperidone ¹⁵ . These findings call for an investigation of how a similar methodology could be used in patients with ADHD.

Despite the importance, there has been minimal examination of this issue. Our group has previously investigated the neural predictors that could suggest stimulant efficacy using diffusion tensor imagery (DTI), finding that structural connectivity in striatal regions correlated with better outcomes ¹⁶ . Other studies have successfully used electroencephalograms (EEGs) to predict stimulant responsiveness, through the recognition of elevated theta power in frontal regions ¹⁷ and slow alpha peak frequencies ¹⁸ . Although the use of DTI and EEG is beneficial, the applicability and practicality are low. A new clinical measure that can be used frequently in real-life clinical settings and at low cost is necessary.

Few studies have investigated clinical predictors of treatment response in ADHD. In a recent systematic review looking into the detection, prediction, and treatment of ADHD, ¹⁹ only 7 out of 100 included trials focused on treatment response, demonstrating a field that needs to be explored. Identifying effective ways to use clinical measures to predict who better responds to stimulant medications before clinicians begin to prescribe would allow for reduced time and effort in trialing medications and a timelier management of ADHD symptoms for the patient. Thus, the main aim of this study is to identify clinical factors that predict a better response to stimulant medications in adult patients with ADHD.

Methods

Participants

Participants were originally from an imaging study conducted by our group, and detailed methods regarding the full scope of the study have been previously reported ¹⁶ . Our initial sample included 40 medication-naïve adult patients diagnosed with ADHD consecutively after referral to the MGH Adult ADHD Program. However, four participants were excluded due to incomplete follow-up data collection, and thus, the final sample for this study includes 36 patients. Participants were assessed by the study clinicians and confirmed that they had the diagnosis of ADHD by meeting the Diagnostic and Statistical Manual of Mental Disorders, 5th Edition (DSM-V) criteria. Participants were excluded if they had concurrent active psychiatric or neurological comorbidities that required urgent clinical attention or if they had neuroimaging contraindications. Participants with mild symptoms or a history of psychiatric conditions that commonly present in ADHD patients, such as anxiety and depression that did not require urgent clinical attention, were not excluded. After the initial clinical assessment, patients were prescribed stimulant treatment for ADHD at therapeutic doses in the normal clinical range and were naturalistically followed in the clinic for two follow-up assessment visits for an average of 116.1 ± 77.9 days (final follow-up). The study was approved by both the MGH Institutional Review Board and the Committee on the Use of Humans as Experimental Subjects at the Massachusetts Institute of Technology (MIT).

Assessments

Patients referred to the MGH Adult ADHD Program completed a battery of rating scales prior to their initial clinic visit that included: a demographic questionnaire, the Adult ADHD Self-Report Scale (ASRS), the Adult Self-Report (ASR), the Behavior Rating Inventory of Executive Function—Adult Version (BRIEF-A), the Social Responsiveness Scale 2 (SRS-2)—ASR, the Emotional Dysregulation Subscale of the Barkley Current Behavior Scale—Self-Report (CBS-DESR), the Mind Wandering Questionnaire (MWQ), and the Quality of Life Enjoyment and Satisfaction Questionnaire (Q-LES-Q).

The ASRS is an 18-item patient-rated questionnaire to determine the severity of ADHD symptoms^20,21. Subdomain scores (inattentive and hyperactive/impulsive) can range from 0 to 36, and patients with a score of ⩾24 are highly likely to have ADHD.

The ASR is a 126-item self-rated assessment of a wide range of psychiatric syndromes in adults. ²² Raw scores are calculated and used to generate T-scores for eight clinical scales, two composite scales, and one total scale. Three of the ASR clinical scales (Attention Problems, Aggressive Behavior, and Anxious/Depressed) can be combined to create a composite T-score for a profile (AAA-profile) that measures levels of emotional dysregulation ²³ .

The BRIEF-A is a 75-item patient-rated questionnaire to assess an adult’s cognitive, emotional, and behavioral functions within the past month ²⁴ . Raw scores are calculated and used to generate T-scores for nine subscales, two summary index scales, and one scale reflecting overall functioning.

The CBS-DESR is a subset of eight questions from the Current Behavior Scale designated by Barkley as measuring Deficient Emotional Self-Regulation (DESR)^25,26 that asks patients to describe their behavior in the past 6 months. Total scores range from 0 to 24.

The MWQ is a five-item scale that assesses mind-wandering traits with each item rated on a Likert scale of 1 (almost never) to 6 (almost always), ²⁷ with total scores ranging from 5 to 30.

The assessment of quality of life relied on the Q-LES-Q, a self-rated 16-item rating scale to assess enjoyment and satisfaction levels in various areas of daily life ²⁸ . The total score is calculated by summing the first 14 items, and scores can range from 14 to 70.

At all study visits (initial clinical assessment and two follow-up visits), participant ADHD severity was assessed with the NIH clinician-assessed ADHD Clinical Global Impression (CGI) Scale of Severity (CGI-S: 1 minimally ill–7 extremely ill) ²⁹ . At the follow-up visits only, participant ADHD symptom improvement was assessed with the CGI Scale of Improvement (CGI-I: 1 very much improved since the initiation of treatment–7 very much worse since the initiation of treatment) ²⁹ . Additionally, the type of stimulant prescribed (MPH or AMPH) was recorded at each study visit.

The reporting of this study conforms to the STROBE statement (Supplemental Table 1: STROBE Checklist) ³⁰ .

Statistical analyses

Participants were stratified into treatment responders and non-responders based on endpoint Clinical Global Impression—Global Improvement (CGI-I) scores ⩽2 and >2, respectively ³¹ . We conducted bivariate analyses comparing baseline demographic and clinical characteristics between treatment responders and non-responders. Truncated Poisson regression models were used to analyze the ASR measure (clinical scales: lower limit (LL) for truncation = 50, upper limit (UL) for truncation = 90; composite scales: LL = 30, UL = 90), BRIEF (LL = 30, UL = 94), SRS (LL = 30, UL = 90), ADHD Rating Scale-IV (ADHD-RS-IV; UL = 36 for subdomains, UL = 72 for total), MWQ (LL = 5, UL = 30), Barkley DESR scale (UL = 24), and Q-LES-Q (LL = 14, UL = 70). After the bivariate analyses, we performed stepwise logistic regression (backwards selection, p ⩾ 0.05 for removal) starting with a model that included all outcomes with p ⩽ 0.10 from the bivariate analyses to identify the best predictors of responding to treatment. We used a less stringent p value of 0.10 in the bivariate analyses to compile our list of starting predictors in the stepwise model to be more inclusive of potential predictors. We estimated the relative contribution to the odds of treatment response for the best predictors by calculating Nagelkerke pseudo-R ² statistics. Lastly, we assessed the predictive utility of the final model using receiver operating characteristic curve analysis and calculating the area under the curve (AUC) statistic. The AUC statistic ranges from 0 to 1 and represents the probability that a randomly selected responder/non-responder pair is accurately classified. At the study endpoint, rates of switching stimulant formulations and rates of responders among those taking AMPH versus MPH were examined using Pearson’s chi-squared test.

As a sensitivity analysis, we examined the moderating effect of ADHD-RS score on the relationship between treatment response and executive functioning as measured by the BRIEF-GEC. We used the same methodology as above but included a term for the responder status-by-ADHD-RS score interaction in the bivariate analyses and a term for the ADHD-RS-by-BRIEF-GEC interaction in the stepwise model.

We also applied Bonferroni corrections within each domain to account for multiple testing in the bivariate analyses. ADHD symptoms on the ADHD-RS were compared against an alpha of 0.016 (0.05/3 tests), psychopathology as measured by the ASR against an alpha of 0.0045 (0.05/11 tests), executive function as measured by the BRIEF against an alpha of 0.004 (0.05/12 tests), and symptoms of autism as measured by the SRS-2 against an alpha of 0.008 (0.05/6 tests). All other clinical measures (MWQ, Barkley DESR, and Q-LES-Q) were compared against an alpha of 0.05, as they were the only assessments in their domains.

All tests, outside of those with Bonferroni corrections applied, were two-tailed and performed at the 0.05 alpha level using Stata^®. ³² All data are presented as percentages, absolute numbers, or mean ± standard deviation (SD) unless otherwise described.

Results

Study sample

Of the 36 participants, 18 (50%) had CGI-I scores ⩽2 at study endpoint and were considered treatment responders. The remaining 18 (50%) participants had CGI-I scores >2 at endpoint and were considered non-responders.

Demographic and clinical characteristics

There were no significant differences between treatment responders and non-responders in age, sex, race, or type of stimulant prescribed at the baseline study visit (Table 1). Responders had significantly more inattentive and total ADHD symptoms as measured by the ADHD-RS compared to non-responders (Table 1). When examining psychopathology via the ASR, treatment responders were significantly less impaired on the Intrusive, Thought Problems, Withdrawn Problems, and Internalizing Problems scales (Table 1). Conversely, when examining executive functioning with the BRIEF, treatment responders were significantly more impaired on the Working Memory and Task Monitoring scales (Table 1). There were no significant differences between the two groups in autism symptoms, mind wandering, emotional dysregulation, or quality of life as measured by the SRS, MWQ, Barkley DESR scale, and Q-LES-Q, respectively (Table 1).

Table 1.

Comparison of demographic and clinical characteristics of patients who did and did not respond to stimulant treatment for ADHD.

Variable	Non-responders, N = 18	Responders, N = 18	p Value
Demographics
Age	30.2 ± 7.1	33.1 ± 6.5	0.22
Male	7 (39)	8 (44)	0.74
Race			0.74
Asian/Native Hawaiian	1 (6)	2 (11)
Black/African American	4 (22)	2 (11)
White/Caucasian	13 (72)	12 (67)
More than one race	0 (0)	1 (6)
Unknown	0 (0)	1 (6)
Type of stimulant			0.72
Amphetamine	6 (33)	5 (28)
Methylphenidate	12 (67)	13 (72)
ADHD Rating Scale
Inattentive	24.1 ± 6.5	27.6 ± 5.5	0.03
Hyperactive	18.8 ± 7.4	20.8 ± 10.6	0.19
Total	42.9 ± 12.5	48.4 ± 15.7	0.02
ASR
Aggressive behavior	56.3 ± 7.1	56.2 ± 6.2	0.95
Anxious/depressed	63.5 ± 9.5	60.5 ± 8.0	0.20
Attention problems	69.3 ± 9.2	72.6 ± 9.6	0.23
Intrusive	57.2 ± 5.6	53.9 ± 6.9	0.04
Rule-breaking behavior	58.9 ± 7.1	56.2 ± 7.6	0.14
Somatic complaints	58.6 ± 8.9	55.3 ± 4.7	0.07
Thought problems	58.2 ± 6.9	54.3 ± 5.4	0.02
Withdrawn	57.7 ± 10.8	53.6 ± 6.4	0.01
Externalizing problems	56.9 ± 8.5	53.3 ± 10.7	0.14
Internalizing problems	61.7 ± 12.2	56.0 ± 9.5	0.03
Total problems	62.9 ± 10.0	59.5 ± 8.2	0.19
BRIEF
Inhibit	59.2 ± 12.9	63.3 ± 10.8	0.12
Shift	56.9 ± 9.9	61.1 ± 12.6	0.10
Emotional control	51.5 ± 11.8	52.3 ± 11.8	0.73
Self-monitoring	50.8 ± 10.2	52.7 ± 14.0	0.44
Initiate	66.7 ± 11.6	67.6 ± 10.9	0.74
Working memory	73.1 ± 12.3	78.6 ± 11.8	0.04
Planning/organizing	68.3 ± 12.1	72.7 ± 11.5	0.11
Task monitoring	68.2 ± 12.0	76.2 ± 12.4	0.004
Organization of materials	58.9 ± 13.5	63.7 ± 13.7	0.07
BRI	55.3 ± 10.7	58.4 ± 11.1	0.22
MI	69.7 ± 11.7	74.9 ± 11.1	0.06
GEC	64.4 ± 10.7	69.3 ± 10.8	0.07
SRS-2
Awareness	48.8 ± 5.5	45.7 ± 8.0	0.18
Communication	50.1 ± 7.0	52.1 ± 8.3	0.41
Cognition	48.7 ± 6.6	50.6 ± 9.0	0.41
Motivation	57.1 ± 10.6	56.1 ± 9.0	0.69
RRB	52.4 ± 8.6	51.9 ± 10.1	0.84
Total	51.5 ± 6.8	51.9 ± 8.0	0.85
Mind wandering	24.4 ± 4.7	25.2 ± 4.2	0.49
Barkley DESR	6.3 ± 5.2	6.0 ± 4.3	0.74
Q-LES-Q	47.6 ± 6.9	52.1 ± 6.6	0.06

Data presented as mean ± SD or N (%). For all scales, except for the Q-LES-Q, higher scores indicate greater impairment. On the Q-LES-Q, lower scores indicate worse quality of life.

ADHD, attention-deficit/hyperactivity disorder; ASR, Adult Self-Report; BRIEF, Behavior Rating Inventory of Executive Function; DESR, Deficient Emotional Self-Regulation; GEC, Global Executive Composite; MI, Metacognition Index; Q-LES-Q, Quality of Life Enjoyment and Satisfaction Questionnaire; RRB, Restricted/Repetitive Behaviors; SRS-2, Social Responsiveness Scale 2.

When we applied Bonferroni corrections to the bivariate analyses, none of the outcomes that met the significance threshold (p < 0.05) in the original analyses met the new thresholds for significance within their domains.

Predictors of treatment response

All clinical characteristics with p values ⩽0.10 in the bivariate analyses were entered into a stepwise logistic regression model to identify the best predictors of treatment response. The model started with the following predictors: ADHD-RS Inattentive scale, ADHD-RS Total scale; the Intrusive, Somatic Complaints, Thought Problems, Withdrawn Problems, and Internalizing Problems scales of the ASR; the Shift, Working Memory, Task Monitoring, Organization of Materials, Metacognition Index, and Global Executive Composite (GEC) scales of the BRIEF; and Q-LES-Q Total score. After removing insignificant predictors one by one based on the largest p value ⩾0.05, we arrived at a final model that identified BRIEF GEC and Q-LES-Q scores as the best predictors of treatment response (Table 2). More impaired executive functioning and better quality of life were associated with significant increases in the odds of treatment response. Using the Nagelkerke pseudo-R estimate, the amount of variance in the odds of treatment response explained by each predictor was 6.8% for the BRIEF-GEC and 13.4% for the Q-LES-Q. When examined together in the same model, the two variables accounted for 36.3% of the variance in the odds of treatment response. The AUC statistic associated with this two-predictor model was 0.81, indicating good predictive utility.

Table 2.

Multiple logistic regression model predicting response to ADHD treatment from BRIEF-GEC T-score and Q-LES-Q Total score (N = 36).

Variable	Odds ratio	95% CI	p Value
BRIEF-GEC	1.13	(1.02, 1.25)	0.019
Q-LES-Q Total	1.23	(1.05, 1.45)	0.010

Results from stepwise estimation using backward selection. Model p value = 0.003.

ADHD, attention-deficit/hyperactivity disorder; BRIEF, Behavior Rating Inventory of Executive Function; GEC, Global Executive Composite; Q-LES-Q, Quality of Life Enjoyment and Satisfaction Questionnaire.

Treatment characteristics at study endpoint

A significantly greater percentage of non-responders switched stimulant medication formulation by study endpoint compared to responders (56% vs 11%, p = 0.005). At baseline, 11 participants were taking AMPH and 25 participants were taking MPH, whereas at study endpoint, 22 participants were taking AMPH and 14 participants were taking MPH. Seventy-one percent (n = 10/14) of those on MPH at study endpoint were responders versus 36% (n = 8/22) of those on AMPH at endpoint (p = 0.04).

Sensitivity analysis

Our sensitivity analysis revealed a moderating effect of ADHD-RS score on the relationship between responder status and BRIEF-GEC. The interaction term was statistically significant (p = 0.04). When we stratified the analysis by responder status, we found that higher (i.e., more impaired) ADHD-RS scores were significantly associated with higher (i.e., more impaired) BRIEF-GEC scores in both groups (both p values <0.01), but the magnitude of change was greater in the non-responder group. In other words, for every one-point increase in ADHD-RS scale score, the BRIEF-GEC score increased by more in the non-responders than in the responders. However, when we put the interaction term in the selection model, it was removed based on a p value ⩾0.05, and we arrived at the same prediction model as in our primary analysis, indicating that this moderating effect is not significant in the presence of other predictors.

Discussion

Our study aimed to identify clinical indicators for stimulant response in adults who are naïve to ADHD medication treatment and found that executive function (BRIEF GEC) and quality of life (Q-LES-Q) scores were the best predictors. In addition, higher scores on the ASR Thought Problems, Withdrawn Problems, Internalizing Problems, and Intrusive Thoughts subscales were also statistically indicative of lower stimulant efficacy. Similarly, poorer working memory and task monitoring identified on BRIEF predicted better stimulant response. These clinical indicators were derived from questionnaires conducted by clinicians and patients, which could be applied easily in clinical settings due to their quick completion and low cost.

Our findings suggested that higher executive functioning impairment measured by BRIEF was significantly associated with better response to stimulant treatment. Of the executive function elements measured, Task Monitoring had the most statistically significant difference between responders and non-responders, where those with higher scores on the scale, or worse task monitoring, showed better response to stimulant treatment. Task monitoring measures the ability to assess one’s own performance during or shortly after a task to ensure accuracy or completion of the task. This finding suggests that task monitoring may be a specific area that impacts stimulant treatment that could lead to more overall improvement of ADHD symptoms.

Working Memory was another executive functioning element associated with stimulant treatment response, where those with higher scores on the scale, or worse working memory, showed better response to stimulant treatment. Working memory measures the ability to store and manipulate information to complete a task. ADHD has been associated with very significant impairment in working memory in both children and adults^33,34. Current literature has suggested that treatment with stimulant medications can enhance aspects of memory in patients with ADHD,^35,36 A neuroimaging study further established this relationship by linking stimulant medication to improved strength of connectivity of some frontoparietal regions. The connectivity changes were directly related to improved working memory in a reaction time task ³⁷ .

While deficits in executive functioning such as organization, initiation, and working memory are separate symptoms from the core symptoms of ADHD such as inattention or hyperactivity, ²⁴ they commonly present and have significant impact on daily functioning in adults with ADHD putting them at risk for lower academic and occupational success ³⁸ . Previous research has shown that while stimulant treatment could have positive short-term impact on executive memory, reaction time, reaction time variability, and response inhibition in individuals with ADHD,^4,39,40 such treatment is not associated with long-term improved executive functioning ⁴¹ . It is interesting that while the direct impact on executive functioning may be inconsistent, executive functioning itself could be a predictor of treatment response to stimulants, indicating an area for future research.

Patients with stronger executive skills often rely more on compensatory strategies to manage ADHD symptoms. Their stimulant response may appear smaller because they already function near their “ceiling.” Those with weaker executive skills lack such compensatory scaffolding, so stimulant-induced improvements are more easily detectable in behavior and performance. Biologically, stimulants primarily enhance dopaminergic and noradrenergic signaling in prefrontal–striatal circuits. Patients with more impaired executive functioning may reflect a greater baseline deficit in these catecholamine systems, meaning there is more room for stimulants to normalize function.

We also found that a better quality of life (QOL) was associated with higher odds of treatment response. Previous literature has almost uniformly indicated that a better baseline quality of life is an indicator for better prognosis for various medical and psychiatric conditions, including cancer ⁴² and schizophrenic spectrum disorders ⁴³ . Higher QOL is often associated with a stable home, school, family support, routines, and resources, which frequently helps with treatment adherence as well as engagement and magnifies the functional impact of stimulant-related symptom reduction. Lower QOL may also reflect overlapping stressors or comorbidities that dilute the apparent medication response. This suggests the importance of support in daily life to improve educational, interpersonal, and social well-being regardless of one’s diagnosis. This finding is also of note as quality of life is often negatively impacted in adults with ADHD. Moreover, additional comorbidities frequently associated with ADHD increase the negative impact on patients’ quality of life^44,45. Medication treatment that leads to symptom improvement in ADHD has also been shown to improve quality of life ⁴⁶ .

Elevated thought problems, withdrawn problems, internalizing problems, and intrusive problems, identified through ASR, also predicted worse treatment response. While these ASR scores were higher in the non-responder group, the elevation was within 1 SD of the norm. This suggests that subthreshold difficulties in these domains could represent a predictor for worse outcomes when using stimulants. The aggregate elevation of thought problems and withdrawn problems, along with social problems comprise an ASD profile ⁴⁷ . The fact that two of the three scales of the ASD profile were linked to worse responses to stimulants could suggest that elements of autism traits could predict worse responses to stimulants in adults with ADHD. This needs further investigation, as our results documented that the SRS, commonly used to identify ASD, did not indicate treatment response in our results.

At endpoint, our study also documented that a significantly greater percentage of non-responders switched stimulant medication formulation from AMPH to MPH or MPH to APH compared to responders. The switches were likely made by the clinicians with the hope that the other formulation would be more effective than the original formulation, which was not effective enough. In our study, there was a larger proportion of responders on MPH than those on AMPH at endpoint. Previous studies have overall documented that the effectiveness of MPH and AMPH are largely similar ⁴⁸ .

Precision medicine allows for a reduction in the trial-and-error process of clinical treatment, achieving better outcomes sooner for patients. ADHD stimulant response has previously been predicted using DTI ¹⁶ and EEG^17,18; however, these processes are time-consuming and costly, making them impractical to use clinically. All of the scales used in our study are typically used in medical practice; therefore, our results can be used in real-life clinical settings as an indicator of stimulant response.

Limitations

The findings in this study are subject to methodological limitations. Despite the value of our sample as a naturalistic clinical sample of medication-naïve adults with ADHD, we had a small sample size for comparison. With only 18 participants in each group of responders versus non-responders, the statistical power of the study is limited. In addition, while all participants received stimulants in therapeutic doses in the normal clinical range, the dose information was not analyzed in this study. Additionally, our sample is from a clinic with a very uniform patient population in race and ethnicity, and thus, the findings may not be generalizable to other populations. The findings would be strengthened if they were replicable in a larger and more diverse patient population across a variety of clinic settings. Several clinical, behavioral, and biological factors that we did not examine in this study may serve as predictors. These factors should be investigated in future studies. Similarly, we did not have clinical diagnoses of depression to incorporate into our statistical models as a potential confounder. Future studies would benefit from including this as a covariate, given that depression can impact executive function and other cognitive domains.

Conclusion

Overall, higher impairment in executive functioning and better quality of life were the best predictors of treatment response. Thought problems, withdrawn problems, internalizing problems, and intrusive problems were also indicative of lower stimulant benefit. Further research in predictors of treatment response in ADHD populations will allow for more precise medication management and a better understanding of variances in best treatment options for prescribing clinicians.