Describing diversity of real world data sources in pharmacoepidemiologic studies: The DIVERSE scoping review.

PURPOSE: Real-world evidence (RWE) is increasingly used for medical regulatory decisions, yet concerns persist regarding its reproducibility and hence validity. This study addresses reproducibility challenges associated with diversity across real-world data sources (RWDS) repurposed for secondary use in pharmacoepidemiologic studies. Our aims were to identify, describe and characterize practices, recommendations and tools for collecting and reporting diversity across RWDSs, and explore how leveraging diversity could improve the quality of evidence. METHODS: In a preliminary phase, keywords for a literature search and selection tool were designed using a set of documents considered to be key by the coauthors. Next, a systematic search was conducted up to December 2021. The resulting documents were screened based on titles and abstracts, then based on full texts using the selection tool. Selected documents were reviewed to extract information on topics related to collecting and reporting RWDS diversity. A content analysis of the topics identified explicit and latent themes. RESULTS: Across the 91 selected documents, 12 topics were identified: 9 dimensions used to describe RWDS (organization accessing the data source, data originator, prompt, inclusion of population, content, data dictionary, time span, healthcare system and culture, and data quality), tools to summarize such dimensions, challenges, and opportunities arising from diversity. Thirty-six themes were identified within the dimensions. Opportunities arising from data diversity included multiple imputation and standardization. CONCLUSIONS: The dimensions identified across a large number of publications lay the foundation for formal guidance on reporting diversity of data sources to facilitate interpretation and enhance replicability and validity of RWE.

Overview of global real-world data sources for pediatric pharmacoepidemiologic research.

PURPOSE: Given limited information available on real-world data (RWD) sources with pediatric populations, this study describes features of globally available RWD sources for pediatric pharmacoepidemiologic research. METHODS: An online questionnaire about pediatric RWD sources and their attributes and capabilities was completed by members and affiliates of the International Society for Pharmacoepidemiology and representatives of nominated databases. All responses were verified by database representatives and summarized. RESULTS: Of 93 RWD sources identified, 55 unique pediatric RWD sources were verified, including data from Europe (47%), United States (38%), multiregion (7%), Asia-Pacific (5%), and South America (2%). Most databases had nationwide coverage (82%), contained electronic health/medical records (47%) and/or administrative claims data (42%) and were linkable to other databases (65%). Most (71%) had limited outside access (e.g., by approval or through local collaborators); only 10 (18%) databases were publicly available. Six databases (11%) reported having >20 million pediatric observations. Most (91%) included children of all ages (birth until 18th birthday) and contained outpatient medication data (93%), while half (49%) contained inpatient medication data. Many databases captured vaccine information for children (71%), and one-third had regularly updated data on pediatric height (31%) and weight (33%). Other pediatric data attributes captured include diagnoses and comorbidities (89%), lab results (58%), vital signs (55%), devices (55%), imaging results (42%), narrative patient histories (35%), and genetic/biomarker data (22%). CONCLUSIONS: This study provides an overview with key details about diverse databases that allow researchers to identify fit-for-purpose RWD sources suitable for pediatric pharmacoepidemiologic research.

Real-world prevalence of homologous recombination repair mutations in advanced prostate cancer: an analysis of two clinico-genomic databases.

BACKGROUND: Homologous recombination repair mutation (HRRm) status may guide risk-stratification and treatment decisions, including polyadenosine diphosphate-ribose polymerase inhibitor use, in advanced prostate cancer. Although HRRm prevalence has been reported in single-institution studies or clinical trials, real-world HRRm prevalence in diverse populations is unknown. We describe HRRm in the clinical setting using two real-world clinicogenomic databases: the Flatiron Health and Foundation Medicine, Inc. Clinico-Genomic Database (CGDB), a national electronic health record-derived database, and the American Association for Cancer Research Project Genomics Evidence Neoplasia Information Exchange (GENIE). METHODS: This cross-sectional analysis included 3757 individuals diagnosed with prostate cancer who had next generation sequencing (NGS) as standard of care. The CGDB included men with advanced/metastatic prostate cancer and genetic data included both germline and somatic pathogenic mutations. The GENIE analysis included men with prostate cancer whose received NGS as standard of care, but the data were filtered to include somatic mutations only. Due to key differences among databases, direct comparisons were not possible. Overall prevalence of HRRm was calculated and stratified by demographic and clinical characteristics. RESULTS: HRRm prevalence (combined germline and somatic) in CGDB (n = 487) was 24.6% (95% CI 20.9-28.7%), with no major differences across demographic and disease characteristic subgroups. HRRm prevalence (somatic) in GENIE (n = 3270) was 11.0% (95% CI 10.0-12.1%), which varied between 9.5% and 18.4% across treatment centers. CONCLUSIONS: Approximately one-quarter of patients with advanced/metastatic prostate cancer in the CGDB had germline and/or somatic HRRm, which is consistent with clinical trials such as the PROfound study that used a similar NGS platform and algorithm to define HRRm. In the GENIE database, HRRm prevalence varied by treatment center or NGS platform. More research is needed to understand real-world HRRm prevalence variations.

Temporal changes in treatment patterns by age group and functional status before and after PD-1/L1 inhibitor approvals in advanced urothelial carcinoma.

Introduction: Metastatic urothelial carcinoma (mUC) has poor prognosis. A high unmet need exists for novel treatment for those who are unfit for platinum-based chemotherapy. Methods: We aimed to describe real-world temporal changes in patient characteristics and 1L treatment selection for mUC patients in the United States following the approval of anti-PD-1/L1 treatments. This study was a retrospective, observational study using anonymized and structured oncology electronic medical record (EMR) data from IQVIA and the US Oncology Network iKnowMed (USON). Results: After approval of 1L anti-PD-1/L1 treatment for mUC, there is a marked increase in the use of 1L anti-PD-1/L1 monotherapies, accompanied by a proportional decrease in 1L platinum-based treatments and non-guideline-based therapy; particularly among the elderly (> 75 years) and those with poor ECOG performance status (ECOG PS 2+). Discussion: Anti-PD-1/L1 monotherapies fulfill the prior unmet need of frail mUC patients who are ineligible for platinum-based therapies.

Assessing Heterogeneity of Treatment Effect in Real-World Data.

Increasing availability of real-world data (RWD) generated from patient care enables the generation of evidence to inform clinical decisions for subpopulations of patients and perhaps even individuals. There is growing opportunity to identify important heterogeneity of treatment effects (HTE) in these subgroups. Thus, HTE is relevant to all with interest in patients' responses to interventions, including regulators who must make decisions about products when signals of harms arise postapproval and payers who make coverage decisions based on expected net benefit to their beneficiaries. Prior work discussed HTE in randomized studies. Here, we address methodological considerations when investigating HTE in observational studies. We propose 4 primary goals of HTE analyses and the corresponding approaches in the context of RWD: to confirm subgroup effects, to describe the magnitude of HTE, to discover clinically important subgroups, and to predict individual effects. We discuss other possible goals including exploring prognostic score- and propensity score-based treatment effects, and testing the transportability of trial results to populations different from trial participants. Finally, we outline methodological needs for enhancing real-world HTE analysis.

Developing Patient-Centered Real-World Evidence: Emerging Methods Recommendations From a Consensus Process.

OBJECTIVES: The Joint ISPOR-ISPE Special Task Force on Real-World Evidence included patient/stakeholder engagement as a recommended good procedural practice when designing, conducting, and disseminating real-world evidence (RWE). However, there are no guidelines describing how patient experience data (PED) can be applied when designing real-world data (RWD) studies. This article describes development of consensus recommendations to guide researchers in applying PED to develop patient-centered RWE. METHODS: A multidisciplinary advisory board, identified through recommendations of collaborators, was established to guide development of recommendations. Semistructured interviews were conducted to identify how experienced RWD researchers (n = 15) would apply PED when designing a hypothetical RWD study. Transcripts were analyzed and emerging themes developed into preliminary methods recommendations. An eDelphi survey (n = 26) was conducted to refine/develop consensus on the draft recommendations. RESULTS: We identified 13 recommendations for incorporating PED throughout the design, conduct, and translation of RWE. The recommendations encompass themes related to the development of a patient-centered research question, designing a study, disseminating RWE, and general considerations. For example, consider how patient input can inform population/subgroups, comparators, and study period. Researchers can leverage existing information describing PED and may be able to apply those insights to studies relying on traditional RWD sources and/or patient registries. CONCLUSIONS: Applying these emerging recommendations may improve the patient centricity of RWE through improved relevance of RWE to patient communities of interest and foster greater multidisciplinary participation and transparency in RWD research. As researchers gather experience by applying the methods recommendations, further refinement of these consensus recommendations may lead to "best practices."

Enhancing Patient Centricity of Real-World Data Research: An Exploratory Analysis Using the Patient Experience Mapping Toolbox.

OBJECTIVES: There is an increasing expectation that medical product development and assessment be guided by patient input captured through patient engagement. Recently published consensus guidelines describe how qualitative patient experience data (PED) can guide the design, conduct, and translation of real-world research that reflects patients' lived experience. The objective of this exploratory analysis is to examine how researchers could leverage PED captured through the Patient Experience Mapping Toolbox (PEMT) to guide real-world data (RWD) research designs. METHODS: This exploratory analysis included a thematic analysis of interview transcripts collected while pilot testing the PEMT followed by a qualitative analysis of the emerging themes aligned with stages listed in the patient-centered real-world evidence, Real-World Research Design Framework. RESULTS: PED collected using the PEMT include information about symptomology, interactions with the healthcare system, information-seeking behavior, misdiagnoses, lifestyle changes, treatments, side effects, and comorbidities. This information can be leveraged at key study design decisions, including (1) identifying study cohorts and subgroups, (2) identifying exposures, (3) informing covariates and potential confounders; and (4) refining study periods. Additionally, participants described where they seek information about treatments and diseases, which should inform dissemination strategies. CONCLUSIONS: We identified opportunities for PED collected using the PEMT to inform RWD study designs. The PED described in this exploratory analysis stem from pilot testing of the PEMT across a variety of conditions. In the next phase of development in this area, researchers should evaluate how data collected using the PEMT can be applied to RWD research for a specific disease.

Noninterventional studies in the COVID-19 era: methodological considerations for study design and analysis.

The global COVID-19 pandemic has generated enormous morbidity and mortality, as well as large health system disruptions including changes in use of prescription medications, outpatient encounters, emergency department admissions, and hospitalizations. These pandemic-related disruptions are reflected in real-world data derived from electronic medical records, administrative claims, disease or medication registries, and mobile devices. We discuss how pandemic-related disruptions in healthcare utilization may impact the conduct of noninterventional studies designed to characterize the utilization and estimate the effects of medical interventions on health-related outcomes. Using hypothetical studies, we highlight consequences that the pandemic may have on study design elements including participant selection and ascertainment of exposures, outcomes, and covariates. We discuss the implications of these pandemic-related disruptions on possible threats to external validity (participant selection) and internal validity (for example, confounding, selection bias, missing data bias). These concerns may be amplified in populations disproportionately impacted by COVID-19, such as racial/ethnic minorities, rural residents, or people experiencing poverty. We propose a general framework for researchers to carefully consider during the design and analysis of noninterventional studies that use real-world data from the COVID-19 era.

Molecular analysis and favorable clinical outcomes in real-world patients with metastatic renal cell carcinoma.

BACKGROUND: Prior biomarker studies have mainly been restricted to advanced RCC patients treated in clinical trials or have had limited integration of immunotherapy features such as programmed death ligand (PD-L)-1 with gene expression signatures intended to capture other canonical pathways to confirm their prognostic value. MATERIAL AND METHODS: PD-L1 and PD-L2 by immunohistochemistry (IHC), PD-L2 messenger RNA (mRNA), and 10 gene expression profile (GEP) signatures targeting immune, angiogenesis and canonical pathways were analyzed in nephrectomy specimens from 227 advanced clear cell RCC (ccRCC) and 42 non-clear cell RCC (nccRCC) patients treated with targeted therapies including VEGF and mTOR inhibitors. Biomarker association with best overall response (BOR), progression-free survival (PFS), and overall survival (OS) were evaluated using multivariable modeling. Except for PD-L1 IHC and angiogenesis, tested with a nominal p-value of .05, multiplicity control was applied with a 0.1 significance level given limited experience in this setting. RESULTS: The strongest biomarker correlations were observed for hypoxia inducible factor (HIF)-2a and angiogenesis signatures (rho = 0.860 [ccRCC], 0.819 [nccRCC]); hypoxia and glycolysis signatures (rho = 0.943 [ccRCC], 0.973 [nccRCC]); PD-L2 mRNA and T-cell-inflamed GEP signatures (rho = 0.764 [ccRCC], 0.897 [nccRCC]); and PD-L2 mRNA and monocytic myeloid-derived suppressor cell signature (rho = 0.787 [ccRCC], 0.815 [nccRCC]). For ccRCC, higher angiogenesis expression was associated with improved BOR (OR:2.85 [95%CI:1.37, 5.93]), longer PFS (HR:0.61 [95%CI:0.45, 0.82]) and OS (HR:0.74 [95%CI:0.54, 1.00]); higher PD-L1 expression with shorter OS (HR:1.44 [95%CI:1.01, 2.07]). For nccRCC, there was more than a two-fold increased risk with longer OS associated with lower angiogenesis (HR:2.43 [95%CI:1.04, 5.68]), glycolysis (HR:7.03 [95%CI:1.51, 32.76]) and hypoxia (HR:8.83 [95%CI:1.69, 46.05]) gene signature expression. CONCLUSION: Data pointed at PD-L1 IHC and angiogenesis expression in ccRCC and hypoxia, glycolysis, and angiogenesis expression in nccRCC as potential prognostic factors. These findings may have implications for the design and interpretation of advanced RCC trials and to identify potential targets for combination therapy strategies.

Regulator-Requested Non-Interventional Postauthorization Safety and Effectiveness Studies for Oncology Drugs: A Systematic Review.

There has been a growing number of oncology drug approvals. Non-interventional postauthorization safety/effectives studies (PASSs/PAESs) aim to provide real-world evidence on the safety/effectiveness of oncology drugs postapproval. To understand the current landscape, a comprehensive search as of March 1, 2021, was conducted in major register/databases. We found that requested studies increased from 1 in 2006-2010 to 47 in 2016-2020. Of 78 total studies identified, 50 focused on safety/risk assessment (64.1%), 6 on effectiveness (7.7%), 3 on drug utilization (3.8%), 1 on disease epidemiology (1.3%), and 18 on effectiveness of additional risk minimization measures (23.1%). For safety/risk assessment studies, 58.9% focused on nonspecific end points (e.g., frequency of adverse events). For effectiveness studies, the leading primary outcome was overall survival. Overall, safety/risk assessment studies concerning nonspecific end points for vascular endothelial growth factor (receptor) inhibitors were most requested. Regarding data sources, though a majority (71.8%) used primary data collection, a growing proportion utilized secondary data sources. Among 23 studies with information available on study design, 10 (43.5%) were single-arm cohort studies, 9 (39.1%) were cross-sectional studies, 3 (13.1%) were comparative cohort studies, and 1 (4.3%) was a nested-case-control study. In conclusion, there was an increasing number of oncology-specific non-interventional PASSs/PAESs, with a majority on safety/risk assessment. Although most utilized primary data collection, there was an increasing use of secondary real-world data sources. Few conducted comparative analyses, and most used single-arm designs. Future efforts are needed to assess how findings of these studies would impact regulatory decisions and improve knowledge of toxicity for clinical/translational development.

A Framework for Extension Studies Using Real-World Data to Examine Long-Term Safety and Effectiveness.

Understanding the long-term benefits and risks of treatments, devices, and vaccines is critically important for individual- and population-level healthcare decision-making. Extension studies, or 'roll-over studies,' are studies that allow for patients participating in a parent clinical trial to 'roll-over' into a subsequent related study to continue to observe and measure long-term safety, tolerability, and/or effectiveness. These designs are not new and are often used as an approach to satisfy regulatory post-approval safety requirements. However, designs using traditional clinical trial infrastructure can be expensive and burdensome to conduct, particularly, when following patients for many years post trial completion. Given the increasing availability and access of real-world data (RWD) sources, direct-to-patient technologies, and novel real-world study designs, there are more cost-efficient approaches to conducting extension studies while assessing important long-term outcomes. Here, we describe various fit-for-purpose design options for extension studies, discuss related methodological considerations, and provide scientific and operational guidance on practices when planning to conduct an extension study using RWD. This manuscript is endorsed by the International Society for Pharmacoepidemiology (ISPE).

Real-world evidence to support regulatory decision-making for medicines: Considerations for external control arms.

Randomized clinical trials (RCTs) are the gold standard in producing clinical evidence of efficacy and safety of medical interventions. More recently, a new paradigm is emerging-specifically within the context of preauthorization regulatory decision-making-for some novel uses of real-world evidence (RWE) from a variety of real-world data (RWD) sources to answer certain clinical questions. Traditionally reserved for rare diseases and other special circumstances, external controls (eg, historical controls) are recognized as a possible type of control arm for single-arm trials. However, creating and analyzing an external control arm using RWD can be challenging since design and analytics may not fully control for all systematic differences (biases). Nonetheless, certain biases can be attenuated using appropriate design and analytical approaches. The main objective of this paper is to improve the scientific rigor in the generation of external control arms using RWD. Here we (a) discuss the rationale and regulatory circumstances appropriate for external control arms, (b) define different types of external control arms, and (c) describe study design elements and approaches to mitigate certain biases in external control arms. This manuscript received endorsement from the International Society for Pharmacoepidemiology (ISPE).

Cardiovascular Events Following Treatment Initiation with Atypical Antipsychotic Medications in Publicly Insured U.S. Youth.

OBJECTIVE: To assess the risk of incident cardiovascular events that led to hospitalizations or emergency department visits following atypical antipsychotic (AAP) treatment initiation in youth according to dose, duration of use, and concomitant use of leading psychotropic medication classes. METHODS: We used computerized Medicaid claims to conduct a retrospective cohort study of youth (5-20 years) who initiated AAP treatment. AAP use was operationalized in a time-dependent manner according to current vs. former use, average daily dose (in risperidone dose equivalents), and duration of use. In a secondary analysis, concomitant use of (1) stimulants and (2) serotonin-reuptake inhibitors (SSRI/SNRIs) with AAPs was also assessed. To account for confounding, disease risk score methodology was used in discrete time failure models. RESULTS: There were 74,700 youth who initiated AAP treatment (average follow-up = 24.8 months). During follow-up, the risk of cardiovascular events was significantly greater during current than former AAP use (RR = 1.55, 95% CI = 1.09-2.21). Furthermore, for current users of AAPs, the risk of cardiovascular events intensified with average daily dose (RR = 2.04, 95% CI = 1.11-3.77 for >3.75 mg/day vs. ≤1.25 mg/day). The risk of cardiovascular events did not significantly differ according to duration of AAP use. In AAP-treated youth, concomitant SSRI/SNRI use was associated with an increased risk of cardiovascular events (RR = 1.61, 95% CI = 1.01-2.57). By contrast, stimulant use concomitant with AAPs was not significantly associated with an increased risk of cardiovascular events. CONCLUSIONS: In publicly insured U.S. youth, current AAP use was associated with an increased risk of incident cardiovascular events, which intensified with increasing dose and with concomitant SSRI/SNRI use. Prudent interpretation of these findings suggests that further research is needed to identify youth subpopulations with the greatest risk of developing AAP treatment-emergent cardiovascular events.

Comparing Nurse Practitioner and Physician Prescribing of Psychotropic Medications for Medicaid-Insured Youths.

Objective: To describe psychotropic medication prescribing practices of nurse practitioners (NP) and physicians for Medicaid-insured youths in 2012-2014 in a mid-Atlantic state where NP independent prescribing is authorized. Method: From annual computerized administrative claims data in a mid-Atlantic state, we analyzed 1,034,798 dispensed psychotropic medications prescribed by NPs and physicians for 61,526 continuously enrolled Medicaid-insured youths aged 2-17 years. Demographic and clinical characteristics of psychotropic medication users were compared for youths who received psychotropic medication dispensings by NP-only, physician-only, or by both providers using descriptive statistics and generalized estimating equations. We then characterized psychotropic medication prescribing practices by providers within each specialty. Results: From 2012 to 2014, the number of psychotropic medication dispensings increased from 346,922 to 349,080. There was a 50.9% increase in the proportion of psychotropic medications prescribed by psychiatric NPs (from 5.9% to 8.8%) and a 28.6% proportional increase by non-psychiatric NPs (from 4.9% to 6.3%). By contrast, the proportion of psychotropic medications prescribed by psychiatrists and by non-psychiatric physicians declined (56.9%-53.0% and 32.3%-31.8%, respectively). Youths diagnosed with depression or anxiety were more commonly treated by NP-only than by physician-only (AOR = 1.33, 95% CI = 1.24-1.43), whereas youths with two or more psychiatric comorbidities were significantly more commonly treated by both NP and physician providers (AOR = 1.44, 95% CI = 1.39-1.50). Psychiatric specialists prescribed the bulk of antidepressants (82.0%) and lithium (92.3%), with much lower prescribing by non-psychiatric specialists (18.0% and 7.7%, respectively). Antipsychotic orders originated from psychiatric specialists 7.4 times more than from their non-psychiatric specialty counterparts, whether physician or NP. Conclusions: NPs, relative to physicians, have taken an increasing role in prescribing psychotropic medications for Medicaid-insured youths. The quality of NP prescribing practices deserves further attention.

The Impact of a State Medicaid Peer-Review Authorization Program on Pediatric Use of Antipsychotic Medications.

OBJECTIVE: This cross-sectional study assessed the impact of a peer-review program on the prevalence of pediatric antipsychotic use among Medicaid-insured youths in a Mid-Atlantic state. METHODS: Medicaid claims (2010-2014) were assessed among continuously enrolled youths in the 12 months before and after implementation of peer review. The study identified children ages zero to four preimplementation (N=118,815) and postimplementation (N=121,431), ages five to nine preimplementation (N=98,681) and postimplementation (N=107,872), and ages 10 to 17 preimplementation (N=154,696) and postimplementation (N=161,370). (Age ranges are inclusive of the final number). In each age group, multivariable logistic regression models with generalized estimating equations assessed the change in annual prevalence of antipsychotic use pre- to postimplementation. Use of other leading psychotropic classes and antipsychotic prescribing by medical specialty were also examined. RESULTS: The annual pre- to postimplementation prevalence of antipsychotic use decreased significantly, from .07% to .03% (adjusted odds ratio [AOR]=.41) among children ages zero to four, from 1.57% to .86% (AOR=.54) among those ages five to nine, and from 3.28% to 2.40% (AOR=.72) among those ages 10 to 17. With the exception of alpha-agonist use, which increased postimplementation (AOR=1.30) among those ages zero to four, no clinically significant pre-post change was noted in other leading psychotropic classes among children ages zero to four and 10 to 17. By contrast, postimplementation use of other psychotropic medications decreased among those ages five to nine (AOR=.73). CONCLUSIONS: A state Medicaid peer-review program resulted in decreased antipsychotic use across all age groups, particularly among children younger than ten. No notable substitution of other psychotropic classes for antipsychotics was observed.

A multi-national comparison of antipsychotic drug use in children and adolescents, 2005-2012.

Over the last decades, an increase in antipsychotic (AP) prescribing and a shift from first-generation antipsychotics (FGA) to second-generation antipsychotics (SGA) among youth have been reported. However, most AP prescriptions for youth are off-label, and there are worrying long-term safety data in youth. The objective of this study was to assess multinational trends in AP use among children and adolescents. A repeated cross-sectional design was applied to cohorts from varied sources from Denmark, Germany, the Netherlands, the United Kingdom (UK) and the United States (US) for calendar years 2005/2006-2012. The annual prevalence of AP use was assessed, stratified by age group, sex and subclass (FGA/SGA). The prevalence of AP use increased from 0.78 to 1.03% in the Netherlands' data, from 0.26 to 0.48% in the Danish cohort, from 0.23 to 0.32% in the German cohort, and from 0.1 to 0.14% in the UK cohort. In the US cohort, AP use decreased from 0.94 to 0.79%. In the US cohort, nearly all ATP dispensings were for SGA, while among the European cohorts the proportion of SGA dispensings grew to nearly 75% of all AP dispensings. With the exception of the Netherlands, AP use prevalence was highest in 15-19 year-olds. So, from 2005/6 to 2012, AP use prevalence increased in all youth cohorts from European countries and decreased in the US cohort. SGA were favoured in all countries' cohorts.

Association of Antidepressant Medications With Incident Type 2 Diabetes Among Medicaid-Insured Youths.

Importance: Antidepressants are one of the most commonly prescribed classes of psychotropic medications among US youths. For adults, there is emerging evidence on the increased risk of type 2 diabetes in association with antidepressant use. However, little is known about the antidepressant treatment-emergent risk of type 2 diabetes among youths. Objective: To assess the association between antidepressant use and the risk of incident type 2 diabetes in youths by antidepressant subclass and according to duration of use, cumulative dose, and average daily dose. Design, Setting, and Participants: A retrospective cohort study was conducted using Medicaid claims data from 4 geographically diverse, large states of youths 5 to 20 years of age who initiated antidepressant treatment from January 1, 2005, to December 31, 2009. Exposures: Antidepressant use (selective serotonin reuptake inhibitors [SSRIs] or serotonin-norepinephrine reuptake inhibitors [SNRIs], tricyclic or other cyclic antidepressants, and other antidepressants) was assessed using the following 4 time-varying measures: current or former use, duration of use, cumulative dose, and average daily dose. Main Outcomes and Measures: Incident type 2 diabetes was assessed using discrete-time failure models, adjusting for disease risk score estimated using more than 125 baseline and time-dependent covariates. Results: In this cohort of 119 608 youths aged 5 to 20 years who initiated antidepressant treatment (59 087 female youths and 60 521 male youths; 54.7% between 5 and 14 years of age) with a mean follow-up of 22.8 months, 79 285 [66.3%] had SSRI or SNRI exposure. The risk of type 2 diabetes was significantly greater during current use than former use of SSRIs or SNRIs (absolute risk, 1.29 per 10 000 person-months vs 0.64 per 10 000 person-months; adjusted relative risk [RR], 1.88; 95% CI, 1.34-2.64) and tricyclic or other cyclic antidepressants (absolute risk, 0.89 per 10 000 person-months vs 0.48 per 10 000 person-months; RR, 2.15; 95% CI, 1.06-4.36), but not of other antidepressants (absolute risk, 1.15 per 10 000 person-months vs 1.12 per 10 000 person-months; RR, 0.99; 95% CI, 0.66-1.50). Furthermore, for youths currently using SSRIs or SNRIs, the risk of type 2 diabetes increased with the duration of use (RR, 2.66; 95% CI, 1.45-4.88 for >210 days and RR, 2.56; 95% CI, 1.29-5.08 for 151-210 days compared with 1-90 days) and with the cumulative dose (RR, 2.44; 95% CI, 1.35-4.43 for >4500 mg and RR, 2.17; 95% CI, 1.07-4.40 for 3001-4500 mg compared with 1-1500 mg in fluoxetine hydrochloride dose equivalents). By contrast, neither the duration nor the cumulative dose of other antidepressants was associated with an increased risk of type 2 diabetes. The risk of type 2 diabetes increased significantly with the average daily dose among youths with more than 150 days of SSRI or SNRI use (RR, 2.39; 95% CI, 1.04-5.52 for >15.0 vs ≤15.0 mg/d) but not among youths with 1 to 150 days of SSRI or SNRI use. Conclusions and Relevance: In a large cohort of youths insured by Medicaid, the use of SSRIs or SNRIs-the most commonly used antidepressant subclass-was associated with an increased risk of type 2 diabetes that intensified with increasing duration of use, cumulative dose, and average daily dose.