UCLA

BACKGROUND. Prescription rates of psychotropic drugs to US children, and average duration of use, rose sharply in recent decades. However, little evidence supports their safety in children, whose physiological, psychological, and social development is still occurring. Previous safety literature is meager, especially on long-term use, but points to potentially harmful short- and long-term physiological consequences that require further investigation. METHODS. From research and reference sources, I constructed a database of 587 non-behavioral physiological outcomes described as potential adverse drug reactions (PADRs) to 104 prescribed psychotropics, affecting 29 human physiological systems. Three pediatric clinicians classified each PADR as non-serious or serious per established standards. I linked PADR information, converted to ICD-9 and ICD-10 diagnoses by three pediatric clinician coders, to the Colorado All-Payer Claims Database (APCD) from 2009 to 2018 (N = 1,066,005 members under 18 years), a comprehensive population-based source of public and private health and pharmacy claims data in that US state. I conducted a retrospective cohort study comparing children first prescribed a psychotropic drug between the ages of 6 and 12 (n1 = 42,362) to randomly sampled psychotropic drug-unexposed children in the same age range (n2 = 42,362). Descriptive analyses of the analytic dataset (n = 84,724; mean followup time: 7.7 years, SD = 2.1; mean duration of psychotropic treatment in the exposed: 4.0 years, SD = 2.9) characterized prescribed psychotropic use and PADRs by duration of use, human physiological system, PADR seriousness, age, gender, median household income by member ZIP code, and epilepsy or recurrent seizure diagnosis. Next, 10 mixed effects Cox proportional hazards models related PADR hazard overall (2 models) and in specific physiological systems (8 models) to psychotropic exposure, polypharmacy, and covariates over time. Cox models accounted for within-subject heterogeneity using a frailty term Z, and included a time-varying predictor for psychotropic exposure. RESULTS. Exposed children experienced substantially more PADRs (n = 582,003) than unexposed children (n = 218,741), with most PADRs in each group occurring in central nervous, sensory, gastrointestinal, and cardiovascular systems. Compared to drug-unexposed APCD members with the same age, gender, income, and seizure disorder diagnosis, psychotropic exposure increased the expected hazard of PADRs of any level of seriousness by 45.0% (HR 1.45, 95% CI [1.39, 1.52]), and serious PADRs by 60.0% (HR = 1.60, 95% CI [1.51, 1.70]). Psychotropic polypharmacy occurred in most (59.2%) of the exposed, and increased the hazard of PADRs compared to children on monotherapy in every statistical model (range of HRs: 1.33, 95% CI [1.10, 1.64], to 7.30, 95% CI [6.29, 8.48]). Exposure interacted significantly with time in every model, suggesting a slowly decreasing expected hazard of PADRs over time in every analysis. To my knowledge, this is the first study of prescribed psychotropic drug harm in children to combine long followup time, multiple drug classes and physiological systems, a population-based dataset, the majority of known or suspected non-behavioral PADRs, and rigorous control for time-varying exposure and within-subject heterogeneity. Its primary limitations include incomplete control of confounding by indication; moderate inter-rater reliability; and absence of control of non-psychotropic prescription drug use.