Risk of Incident Psychosis and Mania With Prescription Amphetamines

Abstract

Objective:

Amphetamine prescribing has increased in the United States in recent years. Previous research identified an increased risk of incident psychosis with prescription amphetamines. The purpose of this study was to examine the impact of dose levels of prescription amphetamines on the risk of this rare but serious adverse outcome.

Methods:

A case-control study using electronic health records was conducted to compare the odds of incident psychosis or mania with past-month exposure to prescription amphetamines. Case subjects were patients ages 16–35 hospitalized at McLean Hospital for incident psychosis or mania between 2005 and 2019. Control subjects were patients with an initial psychiatric hospitalization for other reasons, most commonly depression and/or anxiety. Amphetamine doses were converted to dextroamphetamine equivalents and divided into terciles. Secondary analyses evaluated the odds of psychosis or mania with methylphenidate use.

Results:

Among 1,374 case subjects and 2,748 control subjects, the odds of psychosis and mania were increased for individuals with past-month prescription amphetamine use compared with no use (adjusted odds ratio=2.68, 95% CI=1.90–3.77). A dose-response relationship was observed; high doses of amphetamines (>30 mg dextroamphetamine equivalents) were associated with 5.28-fold increased odds of psychosis or mania. Past-month methylphenidate use was not associated with increased odds of psychosis or mania compared with no use (adjusted odds ratio=0.91, 95% CI=0.54–1.55).

Conclusions:

Although use of hospitalized control subjects excludes individuals with less severe disease, leading to selection bias, the study results suggest that caution should be exercised when prescribing high doses of amphetamines, with regular screening for symptoms of psychosis or mania.

Prescription stimulants are effective treatments for attention deficit hyperactivity disorder (ADHD). While methylphenidate is the predominantly used stimulant in most countries, prescription amphetamines are used at higher rates in the United States (1). Among U.S. adults, a fivefold increase in amphetamine prescriptions occurred between 2004 and 2019, with a further increase of 11%–19% in prescription stimulant use between 2020 and 2021, driven by amphetamines (2, 3).

Neurobiological changes induced by prescription stimulants include increased release of presynaptic dopamine and inhibition of the dopamine transporter. Prescription amphetamines induce a fourfold increase in presynaptic dopamine release compared with methylphenidate (4), whereas methylphenidate is a more potent inhibitor of the dopamine transporter (5). Individuals with psychosis have increased release of dopamine compared with control subjects, and the amount of dopamine released correlates with severity of positive symptoms (6). However, there is no evidence of altered dopamine transporter availability in patients with schizophrenia (7). Therefore, dopaminergic changes induced by amphetamines more closely parallel changes observed with psychosis. In 2006, in response to spontaneous reports of psychosis and mania in children treated with stimulants, the U.S. Food and Drug Administration (FDA) conducted a review of randomized controlled trials (RCTs) and found an increased risk of psychosis and mania among individuals randomized to treatment with stimulants (8). Warnings were added to labels to alert prescribers to the risk of psychosis or mania, which can be serious and may lead to hospitalization (9). A large cohort study found that the risk of psychosis requiring treatment with antipsychotic medications is higher with prescriptions for amphetamines compared with methylphenidate (10). There are currently no published studies to guide prescribers on dose levels and other factors that impact the rare but serious risk of psychosis or mania with prescription stimulants. This represents a major gap in knowledge in light of high rates of prescribing of this class of medications.

Here, we report the results of a case-control study in patients hospitalized for incident psychosis or mania (case subjects) compared with patients with an initial psychiatric hospitalization for other psychiatric reasons (control subjects) to identify dose levels of prescription amphetamines and other factors that may confer an increased risk of psychosis or mania.

METHODS

Data Sources

We used electronic health record data from data repositories with information on all encounters at Mass General Brigham (MGB). MGB is a large health care system that includes two academic medical centers, a freestanding psychiatric hospital (McLean Hospital), and a network of hospitals and outpatient clinics in New England. A unique identifier is assigned to each patient that tracks care across MGB sites. Structured data include encounter dates, ICD diagnosis codes (billing data), demographic characteristics (age, sex, race/ethnicity), insurance plan, medications, and laboratory values. Sex and race/ethnicity were obtained by self-report when individuals register for MGB encounters; self-report of sex did not distinguish between assigned sex at birth or gender identity. Admission notes from McLean Hospital were the primary source of unstructured data. This study was approved by the MGB Institutional Review Board with a waiver of informed consent according to 45 CFR part 46.116 and is reported in accordance with Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for case-control studies (11).

Study Population and Case-Control Definition

Individuals were between the ages of 16 and 35 years and were admitted to McLean Hospital for an initial psychiatric hospitalization between January 1, 2005, and December 31, 2019. Patients on adult inpatient units at McLean Hospital are ≥16 years old; the upper age cutoff was set at 35 years based on the 75th percentile reported for age at onset of schizophrenia (12). Case subjects were admitted for new-onset psychosis or mania. Control subjects had no past or current psychosis or mania and were hospitalized for depression and/or anxiety disorders. Two control subjects were matched to each case subject on admission year. Individuals were excluded if they had a previous hospitalization for psychiatric reasons, previous residential psychiatric treatment, a previous episode of psychosis or mania, or organic CNS disease (see the online supplement).

Patients admitted to McLean Hospital were referred from emergency departments within the MGB system. Upon transfer, patients were initially seen in the centralized Clinical Evaluation Center, where they underwent a psychiatric evaluation that was documented in the admission note; hence, note quality is similar between case and control subjects. To confirm that patients were hospitalized for the first time and were presenting for an initial episode of psychosis or mania, a combination of text mining of notes and structured data was utilized (see Tables S1 and S2 in the online supplement). To ascertain case-control status, text mining was used, where keywords indicative of psychosis or mania (see Table S3 in the online supplement) and surrounding context were extracted every time a keyword appeared in the note. Two psychiatrists with expertise in psychotic disorders manually reviewed phrases (13), classified patients as case or control subjects, and rated their confidence in their classification. Medical notes were reviewed for individuals if at least one clinician was not confident; final classification was decided by consensus. Classification based on phrase review was compared with the gold standard of manual review of medical records associated with admission for 470 randomly selected patients (~10% of patients before control subjects were matched to case subjects). The positive predictive value of classification was 97.0% (see Table S4 in the online supplement).

Exposure

The exposure of interest was past-month prescription amphetamine use: amphetamine/dextroamphetamine (mixed amphetamine salts), dextroamphetamine, or lisdexamfetamine (prodrug of dextroamphetamine). Providers contacted the patient’s pharmacy and documented current medications and dose in the admission note as part of the medication reconciliation process; formulation (extended-release vs. immediate-release) was not consistently reported. Electronic prescribing data were obtained for patients who received outpatient care within MGB. We detected patient-reported misuse of nonprescribed stimulants via text mining for stimulant-related terms (see Table S5 in the online supplement). For admissions between 2017 and 2019, providers accessed Massachusetts’s prescription drug monitoring system, which includes data for controlled substances. (See the online supplement for the hierarchy of methods used for exposure assessment.) We performed a chart review of all patients with prescriptions for stimulants to verify accuracy of recording of exposure, where we noted that all case subjects with prescription amphetamine use were exposed to amphetamines prior to the onset of psychosis or mania.

Prescribed doses were converted to dextroamphetamine equivalents by multiplying daily dose by conversion factors obtained from product labels and reviews by the FDA: 0.75 for mixed amphetamine salts (14) and 0.3 for lisdexamfetamine (15). Distribution of doses across all patients using prescription amphetamines was divided into terciles, corresponding to ≤15 mg dextroamphetamine equivalents (low-dose), >15 mg to 30 mg (medium-dose), and >30 mg (high-dose).

Potential Confounders

Covariates included age at admission, sex, month of admission, race/ethnicity, psychosis or bipolar disorder in a first-degree relative, immigration, insurance type, past-month frequency of cannabis use (daily, two or more times per week, or four or fewer times per month), and binary covariates for alcohol use disorder, current smoking status, and past-month opioid use, cocaine or methamphetamine use, hallucinogen use, or misuse of sedative-hypnotics. Current college attendance was included given higher levels of prescription amphetamine use among college students (16).

We included binary covariates for presence or absence of past psychiatric diagnoses and medications that preceded admission. Past psychiatric diagnoses included ADHD, anxiety disorders, autism, bipolar II disorder, borderline personality disorder, conduct or oppositional defiant disorder, depression, unspecified mood disorders, eating disorders, learning disabilities, obsessive-compulsive disorder, and posttraumatic stress disorder. Psychiatric medications with prescriptions within 60 days of admission included antidepressants, benzodiazepines, methylphenidate, and the non-stimulants guanfacine and atomoxetine. Deemed to be the consequence of outcome, mood stabilizers and antipsychotic medications were not considered confounders and were not adjusted for (17); patients with recent onset of psychosis may have initiated antipsychotic use shortly before admission. Confounders were assessed using structured data (e.g., ICD diagnosis codes, urine toxicology) and text mining for relevant terms in admission notes (see the section on covariate definitions in the online supplement).

Statistical Analysis

Primary outcomes and analysis.

Two primary outcomes were evaluated: the odds of incident psychosis or mania among individuals with past-month amphetamine use versus no use, and the odds of incident psychosis or mania among individuals with past-month use of low-dose, medium-dose, and high-dose amphetamines compared with those without past-month use. Conditional logistic regression models were utilized to assess the odds of psychosis or mania with past-month prescription amphetamine use, adjusted for the prespecified covariates defined above. Logistic regression was utilized for the dose-response analysis; individuals for whom the dose was missing were excluded. Odds ratios with 95% confidence intervals are presented. The attributable risk percentage among those exposed to any prescription amphetamines and to the high-dose level was calculated (18) as (odds ratio−1)/odds ratio.

Effect modification.

Interaction terms were included in separate models to assess whether age, sex, cannabis use, or family history of psychosis or bipolar disorder modified the effect of prescription amphetamines on incident psychosis or mania.

Missing data.

Dose was missing for 53 of the 415 (12.8%) patients exposed to prescription amphetamines. Four other covariates had missing data for 0.6%–1.9% of patients. Multiple imputation was used to address missing data. (See the online supplement for a comparison of methods to address missing dose data.)

Secondary analyses: methylphenidate.

The odds of psychosis or mania with past-month methylphenidate use versus no methylphenidate use were evaluated using conditional logistic regression. In analyses restricted to individuals using prescription stimulants, logistic regression was performed to compare the odds of psychosis or mania between users of prescription amphetamines and methylphenidate, adjusting for covariates and dose. Conversion factors of 0.5 for methylphenidate and 1 for dexmethylphenidate were used to calculate dextroamphetamine equivalents (19, 20),

Sensitivity analyses: selection bias.

Use of hospitalized control subjects can lead to biased estimates (21). Rates of diagnosis of ADHD and prescription amphetamine use are likely higher among patients hospitalized for depression or anxiety than among the intended source population of individuals between the ages of 16 and 35 years who reside in the greater Boston area. Two strategies were used to estimate the effect of selection bias. The first was quantitative bias analysis: In an external validation study in individuals with health care use at MGB, the probability of selection into the study (psychiatric hospitalization) was computed for case and control subjects with and without exposure to amphetamines. We used these parameters to estimate the effect of selection bias on odds ratio for the analysis of past-month amphetamine use versus no use (22) (see the online supplement for details). The second strategy was use of an alternative control group: Case subjects were compared with an outpatient control group consisting of individuals with three or more outpatient visits at the same clinic at MGB for primary care, pediatrics, or psychiatric care. Control subjects were matched 4:1 to case subjects on age (±1 year), sex, race/ethnicity, and date of encounter (±30 days). The same eligibility criteria for the main analysis were applied to outpatient control subjects.

RESULTS

Patient Characteristics

Figure 1 is a flow diagram for the inclusion of 1,374 case subjects with incident psychosis or mania and 2,748 control subjects matched on year of admission. Of the case subjects, 833 (60.6%) were hospitalized for incident psychosis, 388 (28.2%) for incident mania with psychotic features, and 153 (11.1%) for incident mania without psychosis. See Table 1 for the demographic and clinical characteristics of hospitalized case and control subjects. Case subjects were more likely to be male (p<0.001), to be Black (p<0.001), to be Hispanic (p=0.004), and to have public health insurance (p<0.001). Case subjects had higher rates of cannabis use, and daily cannabis use was twice as common in this group (p<0.001). Case subjects were more likely to smoke tobacco (p=0.002) and to report past-month hallucinogen use (p<0.001) but were less likely to report alcohol use disorder (p<0.001), opioid use (p=0.04), and sedative-hypnotic misuse (p=0.005). Prior psychiatric diagnoses and medications in general were more common among control subjects, although case subjects had higher rates of ADHD (p<0.001) and learning disability (p=0.001). Case subjects were more likely to have a family history of bipolar disorder (p=0.03) or psychosis (p<0.001).

FIGURE 1.

Flow diagram of patient inclusion in a study of the risk of incident psychosis and mania with prescription amphetamines

TABLE 1.

Demographic and clinical characteristics of hospitalized case and control subjects and outpatient control subjects^a

Characteristic	Hospitalized Case Subjects (N=1,374)		Hospitalized Control Subjects (N=2,748)		Outpatient Control Subjects (N=5,496)
	Median	IQR	Median	IQR	Median	IQR
Age (years)	22.3	20.3, 26.5	22.5	20.0, 27.5	22.4	20.3, 26.5
	N	%	N	%	N	%
Maleb	893	65.0	1,281	46.6	3,572	65.0
Race/ethnicity
Asian	117	8.5	273	9.9	468	8.5
Blackb	153	11.1	190	6.9	612	11.1
Hispanicb	95	6.9	135	4.9	380	6.9
White	982	71.5	2,080	75.7	3,928	71.5
Other	12	0.9	41	1.5	48	0.9
Not disclosed or unknown	15	1.1	29	1.1	60	1.1
Public insuranceb,c	227	16.5	252	9.2	1,168	21.3
College studentc	552	40.2	1,160	42.2	1,938	35.3
Immigration	226	16.4	400	14.6	887	16.1
Alcohol use disorder (active)b,c	286	20.8	752	27.4	453	8.2
Current smokerb,c	329	23.9	545	19.8	852	15.5
Past-month substance use
Cannabis useb,c					801	14.6
None	668	48.6	1,856	67.5
≤4 times/month	168	12.2	318	11.6
≥2 times/week	143	10.4	183	6.7
Daily	395	28.7	391	14.2
Cocaine or methamphetaminec	83	6.0	129	4.7	135	2.5
Hallucinogenb,c	81	5.9	36	1.3	≤10d	≤0.2
Opioidsb	29	2.1	89	3.2	140	2.5
Sedative-hypnotic misuseb,c	28	2.0	100	3.6	26	0.5
Prior psychiatric diagnoses
ADHDb,c	331	24.1	533	19.4	760	13.8
Anxietyb,c	267	19.4	843	30.7	1,356	24.7
Autism spectrum disorder	18	1.3	52	1.9	58	1.1
Bipolar II disorderc	85	6.2	165	6.0	125	2.3
Borderline personality disorderb	≤10d	≤0.7	31	1.1	14	0.3
Conduct disorder or ODDc	15	1.1	27	1.0	118	2.1
Eating disorderb,c	33	2.4	125	4.5	75	1.4
Learning disabilityb,c	82	6.0	103	3.7	195	3.6
Major depressive disorderb,c	400	29.1	1,705	62.0	747	13.6
Mood disorder, unspecified	33	2.4	66	2.4	116	2.1
Obsessive-compulsive disorderb,c	47	3.4	139	5.1	116	2.1
Posttraumatic stress disorderb	45	3.3	142	5.2	158	2.9
Psychiatric medications on admission (past month)
Prescription amphetaminesb,c	204	14.8	211	7.7	277	5.0
Methylphenidate	37	2.7	102	3.7	159	2.9
Atomoxetine or guanfacine	≤10d	≤0.7	25	0.9	22	0.4
SSRIb,c	226	16.4	977	35.6	664	12.1
SNRIb,c	41	3.0	211	7.7	64	1.2
Bupropionb	48	3.5	257	9.4	158	2.9
Tricyclic antidepressantb,c	≤10d	≤0.7	40	1.5	78	1.4
Other antidepressantb	15	1.1	66	2.4	37	0.7
Benzodiazepineb,c	163	11.9	627	22.8	412	7.5
Antipsychoticb,c	172	12.5	272	9.9	131	2.4
Mood stabilizerb,c	71	5.2	217	7.9	100	1.8
Family history (first-degree relative)
Bipolar disorderb	136	9.9	216	7.9
Psychosisb	73	5.3	47	1.7

^aADHD=attention deficit hyperactivity disorder; IQR=interquartile range; ODD=oppositional defiant disorder; SNRI=serotonin-norepinephrine reuptake inhibitor; SSRI=selective serotonin reuptake inhibitor.

^bSignificant difference between hospitalized case subjects and hospitalized control subjects, p<0.05.

^cSignificant difference between hospitalized case subjects and outpatient control subjects, p<0.05.

^dCell counts ≤10 were suppressed as privacy protection.

Prescription Amphetamine Characteristics

A total of 415 (10.1%) individuals had used prescription amphetamines in the past month. The majority of patients in this group used mixed amphetamine salts (172 control subjects and 175 case subjects, representing 81.5% and 85.8% of control and case subjects, respectively, with past-month amphetamine use). Case subjects were more likely than control subjects to report misuse of nonprescribed stimulants (p=0.007), although most patients had prescriptions for amphetamines (89.1% of control subjects and 79.4% of case subjects).

Primary Analyses

Individuals with past-month prescription amphetamine use had a greater likelihood of psychosis or mania than individuals without past-month use. A dose-response relationship was observed, where higher doses of prescription amphetamine use were associated with greater odds of psychosis or mania. Doses exceeding 30 mg dextroamphetamine equivalents were associated with a 5.28-fold increase in the odds of incident psychosis or mania compared with no amphetamine use (Table 2). The attributable risk percentage was 62.7% for those exposed to any prescription amphetamine, and 81.0% for those exposed to high-dose amphetamines.

TABLE 2.

Primary analyses: odds ratios for past-month amphetamine exposure among hospitalized case subjects compared with hospitalized control subjects

	Hospitalized Control Subjects		Hospitalized Case Subjects		Unadjusted		Adjusted
Measure	N	%	N	%	Odds Ratio	95% CI	Odds Ratio	95% CI
Past-month prescription amphetamine usea	211	7.7	204	14.9	2.14	1.73, 2.63	2.68	1.90, 3.77
Dose responseb
None	2,537	92.3	1,170	85.1	—	—	—	—
Low (≤15 mg)	96	3.5	56	4.1	1.26	0.90, 1.77	1.79	1.15, 2.79
Medium (>15 mg to 30 mg)	64	2.3	61	4.4	2.07	1.45, 2.95	3.51	2.20, 5.61
High (>30 mg)	35	1.3	50	3.6	3.10	2.00, 4.80	5.28	3.06, 9.13

^aAdjusted for age, sex, race, month, insurance type (public or private), immigration, college student, cannabis (frequency), smoking, alcohol, illicit stimulants, hallucinogens, opioids, sedative-hypnotics, attention deficit hyperactivity disorder, anxiety disorder, autism, bipolar II disorder, borderline personality disorder, conduct disorder or oppositional defiant disorder, depression, eating disorder, obsessive-compulsive disorder, posttraumatic stress disorder, unspecified mood disorder, methylphenidate, guanfacine or atomoxetine, selective serotonin reuptake inhibitor, serotonin-norepinephrine reuptake inhibitor, bupropion, tricyclic antidepressant, other antidepressants, benzodiazepines, first-degree relative with bipolar disorder, and first-degree relative with psychosis. Control and case subjects were matched 2:1 on year of admission.

^bAdjusted for the covariates listed above as well as year of admission. Amphetamine dose was converted to dextroamphetamine equivalents; low-, medium-, and high-dose correspond to terciles of amphetamine dose. Fifty-three patients were excluded from the dose-response analysis due to missing dose information, mostly individuals who misused nonprescribed amphetamines.

The online supplement details 20 additional post hoc analyses supportive of the main findings, including, among others, analyses including only individuals with prescriptions for amphetamines, excluding individuals with cocaine or methamphetamine use, and excluding patients taking antipsychotics or mood stabilizers.

Effect Modification

There was significant effect modification by age (p=0.03). Individuals ≤22 years of age (median split) with amphetamine use had lower odds of incident psychosis or mania (adjusted odds ratio=2.26, 95% CI=1.44–3.55) compared with individuals >22 years old (adjusted odds ratio=4.10, 95% CI=2.67–6.32). There was no significant effect modification by sex (females: adjusted odds ratio=2.85, 95% CI=1.73–4.69; males: adjusted odds ratio=3.13, 95% CI=2.09–4.70), cannabis use, family history of psychosis or mania, or any other covariates.

Methylphenidate Analyses

Past-month methylphenidate use was not associated with increased odds of psychosis or mania compared with no use (adjusted odds ratio=0.91, 95% CI=0.54–1.55). Methylphenidate doses were lower than amphetamine doses (mean dextroamphetamine equivalents: methylphenidate, 18.1 mg [SD=12.7]; amphetamine, 25.2 mg [SD=18.3]; p<0.001). In analyses restricted to individuals with past-month stimulant use (N=534), after adjusting for dose and other covariates, prescription amphetamine use was associated with increased odds of psychosis or mania compared with methylphenidate use (adjusted odds ratio=2.85, 95% CI=1.53–5.32).

Sensitivity Analyses

Quantitative bias analysis.

After adjusting for selection bias, the observed odds ratio from the main study increased from 2.13 (unadjusted) to 6.05, suggesting that use of hospitalized control subjects underestimated the effect of amphetamine use on the odds of psychosis or mania. When the analysis restricted control subjects to individuals with depression or anxiety, the odds ratio increased to 4.79 (see the online supplement).

Outpatient control group.

See Table 1 for the demographic and clinical characteristics of 5,496 matched outpatient control subjects. The odds of incident psychosis or mania were greater among those with past-month amphetamine use compared with those with no use. Low-dose amphetamine use was not associated with increased odds of psychosis or mania. Individuals with a daily dose >30 mg dextroamphetamine equivalents had 13.4-fold greater odds of incident psychosis or mania compared with those with no use (Table 3).

TABLE 3.

Sensitivity analysis: odds ratios for past-month amphetamine exposure among hospitalized case subjects compared with outpatient control subjects

	Outpatient Control Subjects		Hospitalized Case Subjects		Unadjusted		Adjusted
Measure	N	%	N	%	Odds Ratio	95% CI	Odds Ratio	95% CI
Past-month prescription amphetamine usea	277	5.0	204	14.9	3.35	2.75, 4.08	2.31	1.71, 3.13
Dose responseb
None	5,219	95.0	1,170	85.1	—	—	—	—
Low (≤15 mg)	159	2.9	56	4.1	1.57	1.15, 2.14	1.01	0.67, 1.52
Medium (>15 mg to 30 mg)	92	1.7	61	4.4	2.96	2.13, 4.11	2.36	1.52, 3.67
High (>30 mg)	18	0.3	50	3.6	12.39	7.20, 21.32	13.38	6.87, 26.05

^aAdjusted for insurance type (public or private), immigration, college student, cannabis (binary), smoking, alcohol, illicit stimulants, hallucinogens, opioids, sedative-hypnotics, attention deficit hyperactivity disorder, anxiety disorder, autism, bipolar II disorder, borderline personality disorder, conduct disorder or oppositional defiant disorder, depression, eating disorder, obsessive-compulsive disorder, posttraumatic stress disorder, unspecified mood disorder, methylphenidate, guanfacine or atomoxetine, selective serotonin reuptake inhibitor, serotonin-norepinephrine reuptake inhibitor, bupropion, tricyclic antidepressant, other antidepressants, and benzodiazepines. Control and case subjects were matched 4:1 on age, sex, race/ethnicity, and date of encounter.

^bAdjusted for the covariates listed above as well as age, sex, race/ethnicity, month, and year of admission. Amphetamine dose was converted to dextroamphetamine equivalents; low-, medium-, and high-dose correspond to terciles of amphetamine dose. Forty-five patients were excluded from the dose-response analysis due to missing dose information.

We observed nonsignificant (p=0.06) effect modification by age, where individuals ≤22 years with amphetamine use had lower odds of incident psychosis or mania (adjusted odds ratio=1.85, 95% CI=1.18–2.89) than older individuals (adjusted odds ratio=3.13, 95% CI=2.12–4.63). There was a significant interaction between prescription amphetamine exposure and sex (p=0.03); females had higher odds of psychosis or mania with amphetamine use (adjusted odds ratio=3.04, 95% CI=1.73–5.32) than males (adjusted odds ratio=2.20, 95% CI=1.56–3.10). There was no evidence of effect modification by cannabis use or other covariates. We identified a nonsignificant (p=0.06) decrease in the odds of incident psychosis or mania among individuals with past-month methylphenidate use compared with those with no use (adjusted odds ratio=0.63, 95% CI=0.39–1.01). In analyses restricted to individuals with past-month stimulant use (N=649), after adjusting for dose and other covariates, prescription amphetamine use was associated with greater odds of psychosis or mania than methylphenidate use (adjusted odds ratio=4.05, 95% CI=2.10–7.80).

DISCUSSION

This case-control study identified an increased odds of incident psychosis or mania with prescription amphetamine use, where increasing dose levels of prescription amphetamines were associated with higher risk. The odds of psychosis or mania with past-month prescription amphetamine use were increased by 5.3 times with doses exceeding 30 mg dextroamphetamine equivalents, which corresponds to 40 mg of mixed amphetamine salts and 100 mg of lisdexamfetamine. In sensitivity analyses comparing case subjects with outpatient control subjects, the odds of psychosis or mania were increased by 13.5 times with the highest dose level. The finding of an increased risk of psychosis with prescription amphetamines compared with methylphenidate is consistent with results from a large cohort study (10).

We found effect modification with age, where individuals older than 22 years of age had higher odds of psychosis or mania than younger individuals (possibly due to higher average amphetamine dose in older group; see the section on sensitivity analyses in the online supplement). Our results suggest that prescription amphetamine dose is the most important modifiable factor associated with stimulant-related psychosis or mania. Current guidelines on ADHD treatment lack maximum doses and recommend targeting dose to symptom control while avoiding intolerable side effects (23), given the lack of evidence-based research supporting maximum doses (24, 25). Our findings suggest that clinicians can mitigate the risk of psychosis or mania by avoiding doses above 30 mg dextroamphetamine equivalents. This dose is higher than the maximum dose for lisdexamfetamine of 70 mg in the FDA-approved label (26), which is equivalent to 21 mg dextroamphetamine equivalents. Most individuals in this study had prescriptions for mixed amphetamine salts; FDA labels do not explicitly provide maximum daily doses for adults (27). High doses of mixed amphetamine salts are commonly prescribed in the United States (28); 11.8% of individuals at MGB in 2019 who had prescriptions for amphetamines had maximum doses above 30 mg dextroamphetamine equivalents (see the online supplement).

This study did not identify an increased risk of psychosis or mania with methylphenidate use. Although some previous work identified an increased risk of psychosis with methylphenidate (29), our findings are aligned with most research suggesting no increase in risk of psychosis with methylphenidate (30–32). A 2-year prospective cohort study in children and adolescents with ADHD with and without methylphenidate use did not identify an increase in psychotic symptoms with methylphenidate use (33, 34).

Meta-analyses of stimulant RCTs have found that in adults, prescription amphetamines were more efficacious and less likely to be discontinued due to adverse effects. However, long-term data are lacking, as the mean duration of exposure to prescription amphetamines in RCTs is 4.8 weeks (35). In a meta-analysis of RCTs focused on dosing strategies (36), flexible titration of amphetamines resulted in improved ADHD symptom control across the range of doses; the authors concluded that doses should be maximized until ADHD symptoms are negligible or intolerable adverse events appear. However, in both meta-analyses, trials where the dose exceeded 30 mg dextroamphetamine equivalents were excluded. Thus, the added benefit of doses beyond 30 mg dextroamphetamine equivalents is unclear. In patients who experience impairment from ADHD symptoms who may benefit from doses above 30 mg dextroamphetamine equivalents, careful monitoring with screening for symptoms of psychosis or mania is critically warranted. This is of particular importance given that psychosis and mania are often associated with lack of insight. Symptoms that develop in the context of prescription amphetamine use could emerge without the patient’s awareness, leading to a delay in detection.

This study has several limitations. Electronic health records are used for routine clinical care and are not intended for research; inconsistencies in documentation and note quality and the use of unstructured text data can lead to misclassification of outcome, exposure, and covariates. We mitigated bias by using a labor-intensive method that included searching through individual notes to minimize missing data. Use of hospitalized control subjects can lead to selection bias (21), as patients hospitalized for depression or anxiety are not representative of the base cohort from which case subjects are derived (in this case, individuals 16–35 years of age who reside in the greater Boston area). This control group excludes individuals with milder depression or anxiety who may possibly benefit from prescription amphetamines. However, findings from sensitivity analyses supported underestimation of the odds of psychosis or mania with prescription amphetamines in the main analyses. This study was not designed for longitudinal follow-up or to capture patient outcomes. Our findings could be due to unmeasured confounders such as higher ADHD severity in case subjects than control subjects. However, the lack of an increased odds of psychosis or mania with methylphenidate use does not support attribution of our results to ADHD severity, although psychiatrists are more likely to start with methylphenidate and switch to amphetamine to target more severe symptoms (37). Reverse causality could be responsible for findings; prodromal symptoms such as cognitive deficits precede the onset of psychosis (38–40) and could lead to increased prescribing of amphetamines or higher doses in case subjects. We lacked consistent data on formulation, onset and duration of amphetamine treatment, and timing of covariates. The effect of prescription amphetamine use on psychosis or mania could be mediated by other factors. We may have inappropriately controlled for factors that are caused by multiple other variables (“colliders”) (41). Our study was conducted in individuals with a first hospitalization at a psychiatric hospital in the greater Boston area, and our findings may not be generalizable to other regions, or to patients with new-onset psychosis or mania who are not hospitalized or who had prior hospitalizations for other reasons, such as depression. The strong associations and dose-response effect observed, a plausible biological mechanism, and consistency with other research (9) support a potential causal relationship between prescription amphetamine use and psychosis or mania. However, this observational study cannot prove causality and is based on a relatively small number of patients on high-dose amphetamines (N=85).

In summary, the results of this study suggest that high doses of prescription amphetamines are associated with an increased odds of incident psychosis or mania. The clinical utility of prescribing doses of amphetamines that exceed 30 mg dextroamphetamine equivalents is unproven; coupled with the elevated risk of inducing psychosis or mania, we recommend minimizing this practice.