Examining daily stimulant medication use and sleep in adolescents with ADHD

Abstract

Research has been inconclusive as to whether stimulant treatment causes or exacerbates sleep problems in adolescents with ADHD. This study examined sleep differences in adolescents with ADHD as a function of stimulant use. Participants were adolescents with ADHD (N = 159, ages 12–14). Parents reported on receipt of stimulant treatment (n = 92, 57.86%; n = 47 amphetamines, n = 45 methylphenidate). Adolescents wore actigraphs and completed daily diaries assessing sleep and daily use of stimulants for 2 weeks. Sleep parameters included daily-reported bedtime, sleep onset latency (SOL), sleep duration, daytime sleepiness, and difficulty waking the following morning; and actigraphy-measured sleep onset time, total time in bed, and sleep efficiency. We estimated between- and within-individual associations between stimulant medication use and sleep indices with all stimulants, after removing adolescents using sleep aids and weekend days, and as a function of stimulant type. Adolescent sleep did not differ between those receiving and not receiving stimulant treatment. Within individuals using stimulants, we largely observed no significant differences between medicated and unmedicated days, though findings were most often significant for school days only. Small effects were found indicating longer SOL, later sleep onset time, and more daytime sleepiness related to medication use. In contrast, there were slight improvements to sleep duration and sleep efficiency related to methylphenidate use, though methylphenidate was also associated with later sleep onset time and more daytime sleepiness. Given the inconsistent and small effects, findings suggest that stimulant medication may impact sleep, but does not appear to be a primary contributor to sleep problems in adolescents with ADHD.

Introduction

Youth with attention-deficit/hyperactivity disorder (ADHD) are more likely to exhibit sleep difficulties, regardless of stimulant-treatment status [1-5], though stimulant treatment may induce or exacerbate difficulties [6-20]. Stimulant treatment has been most robustly linked to longer sleep onset latency (SOL) and shorter sleep duration [9, 10, 12, 14, 17, 19, 21]. Clinical trial data have noted insomnia as an adverse event or side effect [11, 15, 22-24]. Sleep disturbances are typically mild and more likely to occur following increases in dosage, improving as individuals adjust to new medication regiments [9-11, 18]. Other factors that may make sleep disruption more likely include lower weight/body mass index [11], younger age [10, 17], and specific genetic vulnerabilities (e.g., individuals with certain genotype of dopamine transporter gene may be more sensitive to sleep difficulties related to stimulant treatment) [15].

There is also evidence suggesting no association between stimulant treatment and sleep parameters [4, 11, 12, 14, 18, 25-29], or even improved sleep (albeit sometimes following an initial disruption) [4, 18, 25, 27, 30, 31]. Aside from improvements to core ADHD symptoms that may contribute to resistance in going to sleep [4, 6, 31], it is possible that medication may reduce off-task thoughts or behaviors and physiologic hyperarousal (i.e., hyperactivity) that may contribute to sleep problems for the same reasons stimulant medication improves daytime functioning for youth with ADHD [16]. In fact, one might expect that stimulant treatment would improve self-regulation and goal-directed behavior among individuals with ADHD at bedtime, which would lead to less bedtime resistance, fewer off-task thoughts or behaviors, and better attention to and regulation of the body’s natural rhythms for sleep [16]. This is supported by studies finding longer lasting and late-afternoon/early evening doses of stimulant medication to be associated with fewer sleep difficulties [25, 27].

In addition to discrepancies in the literature as to whether stimulant medication truly disrupts sleep in youth with ADHD, several gaps in the literature are worth noting. First, most research has focused on the short-term impact of fixed-dose stimulant treatment among treatment-naïve children [9, 10]. Such findings are important for understanding the impact of medication; however, given findings that point to brief disruptions that are generally improved with adjustment to medication, more research is needed to understand the possible impact of medication use among individuals naturalistically (i.e., individuals on stable, tailored doses). Moreover, treatment-naïve samples are likely less representative of the broader population of youth with ADHD (e.g., less severe, fewer comorbidities) [10]. Second, although research has examined associations within stimulant-treated children, comparing pre- and during-treatment periods, only one study has used the natural variability in adherence to stimulant medication, and its impact on sleep parameters at a daily level [29]. Third, few studies have investigated the impact of stimulant medication on sleep parameters as children transition into adolescence, a period of vulnerability given the common tension between shifts in circadian function in adolescence with the demands of early morning waking for school [10, 32, 33]. Fourth, methylphenidate combinations have been more widely studied than amphetamine combinations [9], which makes sense with respect to this medication being more commonly prescribed in children. However, amphetamines are more commonly prescribed in adolescence (and adulthood) [34]. Additionally, it might be expected for amphetamines to have a stronger negative impact on sleep given they not only block dopamine reuptake, but also increase dopamine release [9, 10].

The present study sought to address discrepancies and gaps in the literature by examining sleep differences measured via both self-report and actigraphy in adolescents with ADHD receiving stimulant treatment with varied treatment adherence over a 2-week period. We also examined sleep differences by general receipt of stimulant treatment, that is, comparing adolescents receiving and not receiving medication treatment (regardless of adherence). Differential effects of methylphenidate and amphetamines were also examined. Given the mixed findings in the literature, we did not have strong a priori hypotheses pertaining to the direction of effects expected, though we expected that any negative impact on sleep indices would be minimal and most likely to be observed with regards to SOL and sleep duration.

Methods

Participants

Participants were eighth grade students (ages 12–14 years; median = 13) recruited from local schools across two sites (Southeastern and Midwestern parts of the United States) as part a larger prospective study (N = 302) in adolescents with and without ADHD [35]. For the current study, we originally included the 162 participants with a diagnosis of ADHD (see below). Because the study focus was on potential influence of stimulant medication on sleep outcomes, we then excluded adolescents that exclusively received non-stimulant medication (n = 3), reducing our sample to 159 adolescents with ADHD. The majority of participants (n = 118, 74.2%) were diagnosed with ADHD predominately inattentive presentation, and the remaining participants (n = 41, 25.8%) were diagnosed with ADHD combined presentation. One-hundred-seventeen (73.6%) adolescents had taken an ADHD medication at some point (initiation ages 3–13 years; median = 8), and 92 (57.9%) were receiving stimulant treatment during the study period (i.e., either stimulant treatment alone or any combination of stimulants and non-stimulants; see Table 1 for more information). Among these participants, 77 (48.4%) were using one stimulant, whereas the rest of the sample was receiving medication treatment with more than one medication (n = 15, 9.4%).

Table 1.

Descriptive statistics by stimulant treatment receipt

	No stimulant tx (n = 66) n (%)	Stimulant tx (n = 93) n (%)	Group differences t, p
Female	18 (27.27)	38 (40.86)	1.77, 0.08
Male	48 (72.73)	55 (59.14)
White	51 (77.27)	76 (81.72)	0.69, 0.49
Non-White	15 (22.73)	17 (18.28)
ADHD presentation
Inattentive	56 (84.85)	62 (66.67)	− 2.62, < .01
Combined	10 (15.15)	31 (33.33)
Any internalizing diagnosis	21 (31.82)	31 (33.33)	0.20, 0.84
Any externalizing diagnosis	16 (24.24)	16 (17.2)	− 1.09, 0.28
Current sleep medication	6 (9.09)	24 (25.81)	2.70,0 < .01
Current SSRI medication	4 (6.06)	13 (13.98)	1.59, 0.11
Current other depression/anxiety medication	0 (0)	3 (3.23)	1.47, 0.14
Current antipsychotic medication	0 (0)	3 (3.23)	1.47, 0.14
Current other medication	1 (1.52)	1 (1.08)	− 24, 0.81
	M (SD)	M (SD)	t, p
ADHD diagnosis age	8.50 (2.30)	7.89 (2.65)	− 1.09, 0.28
ADHD medication initiation age	9.37 (2.06)	8.25 (2.64)	− 1.94, 0.06
Family income	84,394 (32,951)	84,565 (38,026)	0.03, 0.98

Tx treatment. M mean. SD standard deviation. Note: statistically significant group differences are bolded

Procedures

Recruitment for this study took place over the course of a 2-year period (2016–2017), and data used in this study were from the first time point of the longitudinal study: an initial visit and a follow-up visit approximately 2 weeks later. The study was approved by the Virginia Commonwealth University and the Cincinnati Children’s Hospital Medical Center Institutional Review Boards, and we obtained written informed consent and assent. Parents and adolescents meeting initial criteria via parent phone screen were invited to participate in a comprehensive assessment and administration of study measures. Exclusion criteria included the presence of specific adolescent conditions (i.e., organic sleep disorders, autism spectrum disorder, bipolar disorder, or psychotic disorders), adolescent IQ scores below 80 on the Wechsler Abbreviated Scale of Intelligence, Second Edition, homeschooling, and spending the majority of the school day in a special education classroom/resource room. In addition, participants meeting criteria for ADHD predominantly hyperactive-impulsive presentation (n = 2) were excluded given the low prevalence of this presentation in adolescence and ongoing concerns about its validity after early elementary school [36]. At the initial visit, adolescents were given actigraphs to wear over the 2-week period during which time they also completed daily diaries. Follow-up visits were scheduled at least 2 weeks from the initial visit. To be included, families also agreed to maintain the current dosing schedule of any current medications over the duration of the initial study visit and the following 2-week period of data collection.

ADHD and comorbid diagnoses

At the initial visit, parents and adolescents were administered the Children’s Interview for Psychiatric Syndromes [37], which assessed Diagnostic and Statistical Manual for Mental Disorders (DSM), Fourth Edition criteria for mental health disorders, including ADHD; however, criteria were updated for the purposes of our study to reflect DSM-5 criteria (i.e., age of onset of symptom criteria being updated). We used parent report on the interview to determine who met criteria for ADHD, which was defined by meeting all DSM criteria for either ADHD combined presentation (≥ 6 symptoms of inattention and ≥ 6 symptoms of hyperactivity-impulsivity) or ADHD predominately inattentive presentation (≥ 6 symptoms of inattention and ≤ 6 symptoms of hyperactivity-impulsivity). Symptoms of ADHD had to be present prior to 12 years of age; had to be present across settings (e.g., home, school); had to contribute to home, academic, and/or social impairment; and could not be better explained by another mental health disorder. We used combined parent- and adolescent-reported diagnostic interview information to determine whether adolescents also met criteria for other mental health diagnoses (with the exception of assessment for oppositional defiant disorder [ODD], which was only administered to parents; see Table 1). See [35] for additional details.

Measures

Stimulant medication treatment

At the initial visit, parents were administered a modified version of the Services Use in Children and Adolescents Parent Interview (SCA-PI) [38], which provided information on ADHD medication receipt, including (1) whether adolescents had ever received medication management of ADHD, (2) age of ADHD medication initiation, (3) current receipt of medication treatment, and 4) the specific medication name. The daily diary that adolescents completed over the 2-week period asked if they had taken medication for attention concerns that day. Information from the SCA-PI was used for descriptive statistical information and between-individual estimation, whereas the daily adolescent-reported use of medication was used as the primary within-individual predictor for the current study.

Daily sleep diary indices

On the daily diary, adolescents reported (1) how sleepy or tired they felt (not at all [0], just a little [1], pretty much [2], very much [3]) over the course of the day (morning, afternoon, evening), (2) the time they went to bed, (3) how many minutes it took to fall asleep (i.e., SOL), (4) how many hours they slept, and (5) two measures of how tired they were the next morning—how hard it was to wake up (very easy [0], easy [1], neutral [2], difficult [3], very difficult [4]) and how alert they were when they woke up (very alert [0], a little alert [1], neutral [2], a little sleepy [3], very sleepy [4]). Parents and adolescents reported on how hard it was for the adolescent to wake up (very easy [0], easy [1], neutral [2], difficult [3], very difficult [4]). This was the only measure that also utilized parent report. Composite scores were created for daytime sleepiness (i.e., adolescent in the morning, afternoon, and evening) and difficulty waking the next morning (i.e., across adolescents and parents) by summing ratings for each day.

Actigraphy sleep indices

We used AciGraph GT9X Link, worn on adolescents’ nondominant wrist, to measure sleep over the 2-week study period. Data were downloaded using Actilife software version 6. Sixty-second epoch lengths were used to score the actigraph data. They were first validated using both the weartime sensor built into the device and in combination with a validation algorithm to maximize the accuracy of when the actigraph was physically being worn by finding the times of nonwear based on a threshold of consecutive zeros. Once data were validated, sleep scores were calculated with the Sadeh sleep scoring algorithm [39], and by individually adding sleep periods to each night the device was worn for each participant, with adolescent daily diaries used alongside the actigraph scoring. We included the following parameters in the current study: (1) sleep onset time, (2) total time in bed (from sleep onset to offset, documented to be most accurate approximation for total sleep time with actigraphy data in adolescents) [40], and (3) sleep efficiency (SE; defined by total sleep time divided by total time in bed).

Covariates

We included the following between-individual (time-invariant) covariates: any diagnosis of an internalizing disorder, any diagnosis of a disruptive behavior disorder, ADHD presentation, use of other medications, study site, sex, race, and family income. We created indicator variables for any parent- and/or adolescent-reported internalizing disorder (i.e., mood disorders, anxiety disorders, and obsessive compulsive disorder) and any disruptive behavior disorder (i.e., ODD or conduct disorder). We collected information on the receipt of any other prescription medications during the SCA-PI with parents, which was included as a binary indicator variable in models. Study site and demographic variables, including adolescent sex, adolescent race (White, other), and family income (ordinal) were collected during the initial visit. Because so few participants identified as Hispanic but not White (n = 7), we only included race as a covariate in our models. Our only within-individual timevarying covariate was an indicator variable distinguishing between weekend days and school nights, collected through adolescent daily diaries.

Analyses

Data preparation and descriptive statistics were done in SPSS 28 and SAS 9.4. MPlus 8.8 was used for all modeling of primary analyses. We first examined group differences in sample characteristics and sleep indices as a function of stimulant receipt with independent t tests. Next, we computed descriptive statistics for daily medication use and examined mean differences between medicated and unmedicated days with paired sample t tests.

Model estimation aimed to examine both the sleep differences between adolescents with ADHD who were using stimulant treatment for ADHD and adolescents with ADHD who were not using stimulant treatment, as well as the sleep differences within stimulant-treated adolescents as a function of day-by-day stimulant adherence [41, 42]. We examined between-individual associations, because this approach allowed us to understand overall differences in sleep between individuals who did and did not receive stimulant treatment for their ADHD symptoms, and because doing so also allowed us to mirror much of the previous research estimating associations between stimulant treatment and sleep outcomes that use non-treated individuals with ADHD as the comparison. Our primary aim, however, was to compare days of medication use with days of no medication use within the same individuals. Because within-individual estimation uses each person as their own control, time-invariant confounds (e.g., sex, race, genetic factors) are inherently adjusted for. In addition, our daily measurement of both stimulant use and sleep outcomes allowed us to examine whether using one’s stimulant medication negatively impacts their sleep that same day (compared with other days when they did not take their medication), thus providing a more robust test of the study hypothesis. Because we were primarily interested in within-individual variability, we used Dynamic Structural Equation Modeling (DSEM), a Bayesian modeling approach, to examine the stable between-individual and within-individual variation for receipt of stimulant treatment for ADHD and several sleep outcomes over the included 2 weeks of daily diary and actigraphy collection [41, 42]. Bayesian estimation forms a posterior distribution of parameter estimate possibilities based on all available data, and further assumes that the posterior distributions for complete data and for missing data would not differ (the point estimate reported would be the same). Both assumptions are consistent with multi-level handling of missing data assuming missing at random [41, 42]. For the sake of consistency across tests, t tests were also computed with these assumptions (using PROC MIXED in SAS). We had missing data on income (0.63%), daily reports of ADHD medication adherence (0.21%), day of the week of diary completion (5.50%), daily-reported bedtime (9.66%), daily-reported SOL (11.15%), actigraphy-recorded sleep onset time (16.50%), daily-reported sleep duration (11.27%), actigraphy-recorded time in bed (16.50%), actigraphy-recorded sleep efficiency (16.5%), daytime sleepiness in the same day they reported on medication use (9.73%), and difficulty waking the following morning (16.79%). DSEM also assumes stationarity of the data; however, we did observe mean differences by day for the following outcomes: bedtime, SOL, sleep duration, total time in bed, daytime sleepiness, and difficulty waking the next morning. To correct for time trends, we used a step-wise approach to test for and covary patterns by day, using polynomial functions of time (i.e., linear adjustment, quadratic adjustment, etc.) until the time trends (i.e., means and variances) lost statistical significance, the association between our medication and sleep variables lost statistical significance, or the model would not converge.

Models of between- and within-individual associations examined (1) overall associations (regardless of stimulant type), (2) associations after removing adolescents whose parents reported the use of sleep aids over the 2-week period of data collection, 3) associations as a function of stimulant type (amphetamine combinations [n = 47, 30.82%] versus methylphenidate combinations [n = 45, 28.3%]), and (4) associations only including school days, given the likelihood for different sleep patterns and stimulant adherence on the weekend [43, 44].

Results

Variable descriptives by stimulant receipt

There were largely no statistically significant differences in demographic variables observed between those receiving stimulant treatment and those who did not (Table 1). However, adolescents receiving stimulant treatment were more likely to also be taking a sleep aid compared with adolescents not receiving stimulant treatment (t = 2.70, p < 0.01), the most common of which was Melatonin (70% of those using sleep aid were using Melatonin, n = 17 adolescents using stimulants, n = 6 adolescents non using stimulants). In addition, although inattentive presentation of ADHD was most common overall, a larger proportion of adolescents with combined presentation were receiving stimulant treatment (t = − 2.62, p < 0.01).

We present the means for our sleep outcomes in adolescents receiving and not receiving stimulant medication treatment in Table 2; several significant differences emerged. Self-reported bedtime among adolescents receiving stimulant treatment (mean = 10:26 pm) was slightly earlier than adolescents not receiving stimulant treatment (mean = 10:40 pm; t = − 3.04, p < 0.01), and medicated adolescents also reported longer sleep duration (mean for medicated = 8.02; unmedicated = 7.97, t = 2.38, p = 0.02). However, medicated adolescents also self-reported longer average SOL (23.55 min) compared with non-medicated adolescents (18.13 min; t = 4.73, p < 0.01). Adolescents receiving stimulant medication treatment did not differ from adolescents not receiving stimulant medication treatment on other sleep parameters.

Table 2.

Sleep outcomes by stimulant medication receipt

Between individual	No stimulant tx (n = 66)		Stimulant tx (n = 93)		Group differences
	M	SD	M	SD	t, p
Bedtimea	10:40 pm	87 min	10:26 pm	79 min	− 3.04,0 < .01
SOL (minutes)a	18.13	25.84	23.55	33.44	4.73,0 < .01
Sleep onset timeb	23.27	1.68	23.18	1.49	0.28, 0.78
Sleep duration (hours)a	7.97	1.67	8.02	1.78	2.38, 0.02
Total time in bed (minutes)b	483.04	86.83	477.07	94.23	− 1.45, 0.15
Sleep efficiencyb	80.89	7.53	80.36	8.19	− 1.39, 0.17
Daytime sleepinessa	5.48	2.11	5.25	2.00	− .07, 0.94
Difficulty wakinga	5.59	1.84	5.35	1.8	− 3.57, 0 < .01
Within individual	Unmedicated days (n = 777)		Medicated days (n = 1335)		Mean differences
	M	SD	M	SD	t, p
Bedtimea	10:28 pm	76.80	10:28 pm	78.00	2.12, 0.03
SOL (minutes)a	26.59	40.65	23.72	31.65	− 2.16, 0.03
Sleep onset timeb	11:19 pm	87.00	11:16 pm	74.52	0.37, 0.71
Sleep duration (hours)a	8.19	2.08	7.92	1.62	− 2.20, 0.031
Total time in bed (minutes)b	476.83	97.74	477.53	92.01	0.80, 0.42
Sleep efficiencyb	80.17	8.37	80.34	8.14	0.24, 0.81
Daytime sleepinessa	5.02	1.81	5.45	1.84	1.70, 0.09
Difficulty wakinga	5.39	2.00	5.34	2.01	0.46, 0.65

^aSleep indices measured with daily diaries.

^bSleep indices measured with actigraphy. Tx treatment. M mean. SD standard deviation. Note: statistically significant mean differences are bolded

Daily stimulant medication use

When looking across the 2-week period of adolescent daily diary reports, there were 777 reports of medication use total. Daily average medication use was 63.13% among adolescents who reported general receipt of ADHD medication at the initial visit (range: 52.27%-70.45%), with a daily average use of 74.02% on school days. When examining recent stimulant medication initiation and/or dose adjustments to stimulant medication, only 2 adolescents had either initiated their stimulant or adjusted their dose within the previous 30 days of the study, and 4 adolescents adjusted their dose of a stimulant during the 2-week period of collection of diary and actigraphy information.

We report the means and standard deviations for our sleep parameters for both medicated and unmedicated days within individuals receiving stimulant treatment in Table 2, along with paired sample t tests examining mean differences between medicated and unmedicated days. We observed shorter self-reported SOL on medicated days (23.72 min) compared with unmedicated days (26.59 min) by 2.87 min (t = − 2.16, p = 0.03), but also shorter average self-reported sleep duration on medicated days (7.92 h sleep) compared with unmedicated days (8.19 h sleep) by 16.2 min (t = − 2.20, p = 0.031). We did not observe statistically significant mean differences between medicated and unmedicated days for any other sleep indices.

Primary analyses

We present our modeling results grouped by outcome (Table 3). We do not present models without adjustment for covariates and time trends, as results were largely commensurate. When estimating both between- and within-individual variance in models, we only observed one statistically significant association between general stimulant treatment receipt (yes/no) and our self-reported bedtime when we restricted models to only include school days, suggesting that participants receiving stimulant treatment reported earlier bedtimes on school days (b = − 0.43, p < 0.05). As such, we only report on within-individual estimates below (see Table 3 for all estimates).

Table 3.

Adjusted associations between daily medication usage and sleep outcomes

	Bedtimea	SOLa	Sleep onset timeb	Sleep durationa	Total time in bedb	Sleep Efficiencyb	Daytime sleepinessa	Difficulty wakinga
	b (95% CI)	b (95% CI)	b (95% CI)	b (95% CI)	b (95% CI)	b (95% CI)	b (95% CI)	b (95% CI)
Between ind. estimate
Overall	− 0.23 (− 0.53, 0.07)	1.56 (− 2.71, 6.02)	0.05 (− 0.59, 0.42)	0.14 (− 0.22, 0.52)	− 3.33 (− 22.88, 31.12)	− 0.39 (− 1.21, 0.48)	− 0.35 (− 0.94, 0.22)	− 0.21 (− 0.83, 0.38)
Excluding sleep aids	− 0.31 (− 0.71, 0.09)	− 0.76 (− 4.72, 3.19)	− 0.14 (− 0.59, 0.28)	− 0.13 (− 0.61, 0.34)	− 14.89 (− 36.33, 7.29)	− 0.17 (− 3.75, 3.03)	− 0.31 (− 1.06, 0.44)	0.01 (− 0.71, 0.74)
Amphetamine users	− 0.44 (− 0.96, 0.06)	− 1.69 (− 6.00, 2.63)	0.11 (− 0.61, 0.40)	0.11 (− 0.64, 0.85)	− 5.32 (− 36.90, 37.65)	0.60 (− 2.40, 3.70)	− 0.39 (− 1.16, 0.35)	0.52 (− 0.34, 1.29)
Methylphenidate users	− 0.50 (− 1.06, 0.04)	− 0.53 (− 4.65, 3.66)	0.22 (− 0.75, 0.32)	−0.23 (− 0.87, 0.34)	− 26.43 (− 51.59, 1.97)	− 2.19 (− 5.22, 1.09)	− 0.93 (− 1.92, 0.05)	− 0.50 (− 1.29, 0.41)
Excluding weekends	− 0.43 (− 0.85, − 0.00)	0.37 (− 4.64, 5.08)	− 0.02 (− 0.49, 0.44)	− 0.01 (− 0.57, 0.58)	− 11.01 (− 30.59, 8.72)	− 0.40 (− 2.83, 2.18)	− 0.76 (− 1.58, 0.13)	0.08 (− 0.75, 0.95)
Within ind. Estimate
Overall	0.01 (− 0.10, 0.14)	0.00 (− 0.05, 0.05)	0.02 (− 0.13, 0.17)	− 0.20 (− 0.39, −0.01)	− 8.18 (− 22.88, 31.12)	0.42 (− 2.54, 1.83)	0.30 (0.11, 0.47)	0.05 (− 0.04, 0.15)
Excluding sleep aids	0.01 (− 0.01, 0.01)	0.06 (0.03, 0.10)	0.001 (< 0.01, 0.02)	0.00 (− 0.01, 0.01)	− 0.03 (− 0.56, 0.53)	− 0.02 (− 0.06, 0.02)	0.004 (− 0.00, 0.01)	0.00 (− 0.00, 0.01)
Amphetamine users	− 0.00 (-− .01, 0.01)	0.12 (0.08, 0.17)	0.00 (− 0.01, 0.01)	− 0.01 (− 0.02, 0.002)	− 0.12 (− 0.79, 0.59)	− 0.05 (− 0.09, < 0.01)	0.00 (− 0.01, 0.01)	− 0.00 (− 0.01, 0.01)
Methylphenidate users	0.01 (− 0.00, 0.02)	0.00 (− 0.01, 0.01)	0.00 (0< 0.01, 0.02)	0.01 (0< 0.01, 0.02)	0.19 (− 0.34, 0.74)	0.07 (0.02, 0.12)	0.03 (0.02, 0.04)	0.01 (− 0.01, 0.03)
Excluding weekends	0.01 (0< 0.01, 0.01)	0.02 (0.01, 0.04)	0.01 (0< 0.01, 0.01)	0.01 (0< 0.01, 0.01)	− 0.31 (− 0.56, − 0.04)	0.01 (− 0.02, 0.05)	0.01 (0.01, 0.02)	0.00 (− 0.01, 0.01)

^aSleep indices measured with daily diaries.

^bSleep indices measured with actigraphy. Ind individual. SOL sleep onset latency. Credibility intervals that do not include zero indicate statistically significant associations (in bold)

Daily diary-reported bedtime

We did not observe a significant association between medicated and unmedicated days in our overall models examining self-reported bedtime, nor did we observe statistically significant associations when we excluded adolescents using sleep aids or when we examined associations by stimulant type. Similar to between-individual associations, only one statistically significant within-individual association was observed for daily diary-reported bedtime as a function of reported stimulant use: when restricting estimates to school days. However, in contrast to the between-individual results (see preceding paragraph), findings in this case suggested later bedtime on days of stimulant use compared with unmedicated days (b = 0.007, p < 0.01, 95% Credibility Interval [CI] = − 0.85 − 0.001).

Daily diary-reported SOL

In covariate adjusted models estimating the between- and within-individual association between stimulant use and SOL, we did not observe a statistically significant association between medicated and unmedicated days in our overall models examining self-reported SOL (similar to our models above examining self-reported bedtime). We did, however, observe a statistically significant within-individual effect in our subsample excluding adolescents who used sleep aids (b = 0.06, p < 0.001, 95% CI = 0.03–0.10, 95% CI = 0.002–0.01) and among amphetamine users specifically (b = 0.12, p < 0.001, 95% CI = 0.08–0.17) indicating longer SOL on medicated days. We did not observe a statistically significant association for methylphenidate users. Finally, we again observed a significant association when only including school days (b = 0.02, p < 0.001, 95% CI = 0.08–0.17), such that adolescents took longer to fall asleep on days when they reported taking their stimulant medication compared with days when they reported not taking their stimulant medication.

Actigraphy-measured sleep onset time

As with bedtime and SOL, we did not observe a statistically significant association between medicated and unmedicated days in our overall models examining actigraphy-measured sleep onset time. However, similar to estimates of SOL (but not bedtime), adolescents had a later sleep onset time on days they reported taking their medication compared with days when they reported that they did not take their medication in our subsample excluding all adolescents who reported the use of sleep aids (b = 0.007, p = 0.04, 95% CI = < 0.001–0.02). Methylphenidate users, but not amphetamine users, also had a later sleep onset time on medicated days (b = 0.003, p = 0.04, 95% CI = < 0.001–0.02). Finally, as with our other estimates, we observed later sleep onset time on medicated days when only including school days (b = 0.006, p = 0.01, 95% CI = 0.001–0.01).

Daily diary-reported sleep duration

Unlike our previous estimates, we observed a statistically significant association between medicated and unmedicated days related to self-reported sleep duration in our overall estimates of all stimulant users (b = − 0.20, p = 0.04, 95% CI = − 0.39– − 0.01), suggestive of a shorter sleep duration on medicated days. We did not observe a statistically significant association when we excluded adolescents who used sleep aids, nor among amphetamine users. Methylphenidate users specifically had more hours of sleep on medicated days (b = 0.01, p = 0.04, 95% CI = < 0.001–0.02). When looking at only school days among stimulant users, adolescents again reported more hours of sleep on medicated days compared with medicated days (b = 0.01, p = 0.04, 95% CI = < 0.001–0.01), as opposed to fewer hours.

Actigraphy-measured total time in bed

In contrast to our results for sleep duration using daily diary information collected by adolescents, we did not find a significant association in overall estimates among all stimulant users. We also did not observe significant associations between medicated and unmedicated days when excluding adolescents who used sleep aids, nor among amphetamine or methylphenidate users specifically. There was a statistically significant within-individual variation for our models estimating total time in bed when including school days only, with less total time spent in bed on medicated compared to unmedicated school days (b = − 0.31, p = 0.03, 95% CI = − 0.56– − 0.04). This is in direct contrast to selfreports of longer sleep durations and longer SOL on school nights on medicated school days compared to non-medicated school days.

Actigraphy-measured sleep efficiency

We did not observe statistically significant differences in sleep efficiency between medicated and unmedicated days in overall estimates or when excluding adolescents using sleep aids. However, the within-individual analysis with statistical significance showed that methylphenidate users had better sleep efficiency on medicated days compared with unmedicated days (b = 0.07, p = 0.006, 95% CI = 0.02–0.12). We also observed a trending association for amphetamine users that suggested worse sleep efficiency on medicated days compared with unmedicated days (b = − 0.05, p = 0.05, 95% CI = − 0.09–0 < 0.001), thus corroborating the possibility that amphetamines may pose greater risk for poor sleep than methylphenidate. No statistically significant differences were observed when only including school days.

Daily diary-reported daytime sleepiness

We observed statistically significant within-individual associations among all stimulant users (b = 0.30, p < 0.001, 95% CI = 0.11–0.47; school days: b = 0.01, p < 0.001, 95% CI = 0.01–0.02), indicating more sleepiness on the days adolescents reported having taken their stimulant medication. This significant association did not hold when excluding adolescents using sleep aids or among amphetamine users. We did observe a significant association again among methylphenidate users specifically (b = 0.03, p < 0.04, 95% CI = 0.02–0.04) and when only including school days (b = 0.01, p < 0.05, 95% CI = 0.01–0.02), suggestive of more daytime sleepiness on medicated days.

Daily diary-reported difficulty waking

There were no statistically significant within-individual associations between stimulant use and difficulty waking the morning following medication use.

Discussion

The current study used observational data from a sample of adolescents with ADHD to examine between- and within-adolescent differences in several objective and subjective sleep indices as a function of taking stimulant medication. Importantly, adolescents receiving stimulant treatment were largely on stable doses of medication for a longer duration than the samples in most prior studies. We did not observe many differences in sleep parameters as a function of general stimulant receipt, suggesting that observed mean differences in self-reported bedtime and sleep onset time did not have to do with stimulant-treatment status.

When examining the impact of adolescents’ use of medication on sleep parameters at the daily level, the most robust associations were for longer self-reported SOL, later actigraph-measured sleep onset time, and more daytime sleepiness rated three times throughout the same day of reported medication use, consistent with some prior literature [9, 10, 12, 14, 15, 19, 21]. Our finding for daytime sleepiness is most surprising, as one might expect the use of stimulants to help with sleepiness in the same day of stimulant use (but contribute to difficulties falling asleep later that night). These findings may reflect reverse causation of some kind, as individuals may have been more likely to take their medication on days in which they felt more tired and perceived more need for the medication. Although we did not have a measure of perceived need for medication included in the current study, future research would benefit from examining this hypothesis more carefully.

It is important to interpret findings in the context of the number of tests run and whether the significant effects are clinically meaningful. Although it is common for researchers to dedicate more discussion to significant findings, we believe that it is important to stress that in most of our models, we did not observe many sleep differences between medicated and unmedicated days. One might expect convergence in findings across certain indices (e.g., self-reported bedtime, actigraphy-measured sleep onset time, and self-reported SOL), though, similar to the present findings, discrepancies are common [9]. For instance, we observed longer self-reported duration of sleep on school days, yet longer self-reported SOL and shorter actigraphy-measured total time in bed. Moreover, the magnitude of associations observed with statistical significance was also weak, which may explain the lack of convergence across findings. Means for medicated and unmedicated days only significantly differed for SOL and sleep duration, whereby SOL was actually shorter on medicated days than unmedicated days, and sleep duration only differed by 16.2 min indicating shorter sleep duration on medicated days. Other sleep did not show statistically significant mean differences. For instance, medicated and unmedicated days differed by less than 5 min for sleep onset time (with an earlier sleep onset for medicated days), less than a minute for total time in bed, and by 0.17 for sleep efficiency. However, these mean differences were computed among all stimulant users, and the significant findings we did observe were observed most often when we stratified or restricted this group (e.g., looking by stimulant type, looking only among school days). Still, most of the evidence in the present study points to minimal or no impact of stimulant medication on sleep than it does detrimental effects.

It is with this context in mind that we further interpret statistically significant findings. Drilling down on results by analysis, we only observed an impact of stimulant treatment on both self-reported SOL and actigraph-measured sleep onset time when excluding individuals using sleep aids. Findings may provide further support for any stimulant-related difficulties falling asleep being potentially remedied by the use of sleep aids such as melatonin [9, 10, 44], which was the most commonly used sleep aid in our study (accounting for 70% of sleep aid use). This finding is particularly important within the context of other research that has found that the sleep difficulties at baseline are likely to predict both stimulant efficacy and sleep difficulties at stimulant initiation, with mixed findings as to whether sleep difficulties are improved or worsened from baseline [11, 21, 28, 44, 45]. Overall, our findings in conjunction with other research underscore the importance of assessing pre-treatment sleep difficulties to properly gauge any sleep difficulties that may be caused, worsened, or even improved [10], by stimulant treatment.

Most prior research has specifically examined the effects of methylphenidate on sleep [9, 12, 14, 17, 19, 21]. The present study separated out amphetamine and methylphenidate use and found amphetamine use had a negative impact on SOL, consistent with some prior research [9, 10]. In the current study, the only associations with sleep observed related to methylphenidate use were sleep onset time and daytime sleepiness, and we actually observed longer sleep duration and improved sleep efficiency on medicated days. Although seemingly discrepant with prior literature, several studies have observed no impact or even improved sleep related to methylphenidate use [4, 25, 30, 31]. More broadly, one might expect more difficulties falling asleep related to amphetamines given the pharmacodynamic differences between methylphenidate and amphetamine preparations (i.e., amphetamines not only block dopamine reuptake, but also increase dopamine release) [9, 10].

Finally, we observed the most robust pattern of findings indicating worse sleep when we examined associations only among school days. More specifically, medicated school days were related to later self-reported bedtime, longer self-reported SOL, later actigraphy-measured sleep onset time, less actigraphy-measured time spent in bed, and more daytime sleepiness. Given the shifts for later circadian preference that occur in adolescence along with the demand for early morning waking with school [10, 32, 33], these findings may indicate an interaction in need of further investigation. That is, adolescents with later circadian preference than what the demands of their schedule will allow for to get sufficient sleep may be particularly vulnerable to the impact of stimulant-related sleep difficulties.

There are several limitations to the current study that are important to acknowledge and address in future work. First, because of our use of observational data, it is possible that we are under-estimating the negative impact of medication on sleep either because of individuals having reduced dose of medication to manage sleep side effects, or even discontinuation of medication altogether related to sleep side effects. However, even among randomized-controlled trials, such side effects are often minimal and transient, if observed at all [9, 10]. Second, although a strength of our study is that we examined differences between methylphenidate and amphetamine preparations, we did not have reliable information on dose, formulation half-life, route of administration, or schedule of administration, all of which are important for understanding the impact of stimulant medication on sleep parameters. Third, the current sample was largely White and came from higher income families with parents with higher levels of education, which perpetuates problems with research of this type having limited social, racial, and ethnic generalizability. Fourth, although a strength of our study was our use of both sleep diaries and actigraphy, self-report may be biased by problems with recall and missingness (i.e., incomplete information may be more likely on days that medication was missed), and actigraphy is often not sensitive to differences in sleep versus awake states (and thus the transition between states) [46], which is particularly salient given restless sleep is common among youth with ADHD [1]. As such, neither self-report nor actigraphy directly assesses sleep (e.g., sleep architecture and sleep stages). Future studies should examine sleep with polysomnography to directly measure sleep, including sleep structure. Fifth, comprehensive assessment of sleep disorders may be an important avenue for future research to investigate related to possible stimulant-related effects, as this may provide clinically meaningful context to the current literature. Sixth, although the use of repeated measures helps to increase power, our study was likely still limited in power due to our small sample size. As such, our lack of statistically significant findings may be related to this lack of power to detect sleep disruptions due to stimulant treatment. However, this is still consistent with our conclusion that any effects are likely to be small in magnitude. Additional directions for future research include examining possible moderators (e.g., pre-treatment sleep difficulties, ADHD presentation, and age/periods of development) that differentiate those who have stimulant-related sleep difficulties and stimulant-related sleep improvements, as well as mediators of effects (e.g., stimulant treatment reducing bedtime resistance or mind-wandering, thus improving sleep). Research examining perceived need for medication or motivation for medication use at the daily level is also needed in order to test the possibility that individuals may be more likely to use their stimulant medication when they are more tired.

Although the present study found largely minimal-to-no impact of stimulant treatment on sleep in adolescents with ADHD, small effects were found indicating longer SOL, later sleep onset time, and more daytime sleepiness in the same day of medication use. Findings were also most robust for school days, suggesting that stimulant medication use may have more consistent effects on sleep for adolescents with ADHD when their school start times are asynchronous with their circadian rhythms and/or following days of school, which can be challenging for many adolescents with ADHD given their frequent academic and socio-emotional difficulties [47-50]. In contrast, small associations were observed indicating improved sleep duration and sleep efficiency related to methylphenidate use, though methylphenidate was also associated with later sleep onset time and more daytime sleepiness. Overall, the associations suggest that stimulant medication use may have a negative impact on sleep functioning in adolescents with ADHD, especially on school days, which warrants assessment and monitoring of sleep in stimulant-taking adolescents. However, it is important to emphasize that all significant effects were small in magnitude, indicating that stimulant medication is not likely to be a primary cause or contributor to sleep problems in adolescents with ADHD. Research is needed to further investigate factors that contribute, in isolation or in interaction with stimulant medication use, to sleep problems in adolescents with ADHD.

Data availability

Data preparation and descriptive statistics were done in SPSS 28 and SAS 9.4. MPlus 8.8 was used for all modeling. Data and code are available from the corresponding author upon reasonable request and execution of a data use agreement.