A Large-Scale Collaborative Synthesis of Prevention and Treatment Trials for Adolescent Depression: Methodologic, Scientific, and Policy Perspectives

Transcription

1 A Large-Scale Collaborative Synthesis of Prevention and Treatment Trials for Adolescent Depression: Methodologic, Scientific, and Policy Perspectives C Hendricks Brown Department of Psychiatry and Behavioral Sciences Department of Preventive Medicine Institute for Public Health and Medicine Center for Engineering and Health Director, Center for Prevention Implementation Methodology (Ce-PIM) Director, Prevention Science and Methodology Group (PSMG) Hendricks.brown@northwestern.edu

2 In Partnership with Substance Abuse and Mental Health Services Administration (SAMHSA) Acknowledgments Funded by National Institute of Mental Health (NIMH) R01MH : Collaborative Data Synthesis of Randomized Trials for Adolescent Depression National Institute on Drug Abuse (NIDA) and Office of Behavioral and Social Science Research (OBSSR) P30 DA Center for Prevention Implementation Methodology

3 Co-Authors George Howe, George Washington University Tatiana Perrino, Hilda Pantin, Ahnalee Brincks, Shi Huang University of Miami Robert Gibbons, University of Chicago Kwan Hur, VA and University of Chicago William Beardslee, Harvard University Irwin Sandler, Arizona State University Gracelyn Cruden, Juned Siddique, Peter de Chavez, Northwestern University 3

4 Two Fundamental Challenges Guiding our Interests in Improving Behavioral Health for Children, Youth, and Families 1. How do you improve system performance to deliver effective programs for overall benefit? Discovering what interventions work under what conditions: including Prevention Implement Evidence-Based Programs in Ways that Address Local Needs We have numerous effective prevention programs but these are not often used in practice: IOM How do you improve system performance for the most vulnerable, especially when the system itself is poorly functioning? Disparities Health Equities Scientific Equity is needed first. Relevant to the Social Work Grand Challenge: The Coalition for the Promotion of Behavioral Health Social work is ideally positioned to design, deliver, and test programs aimed at preventing behavioral health problems. 4

5 The Concept of Scientific Equity --- Equality in the amount of scientific knowledge that is produced to understand both the causes and solutions to health inequities -- Including the Application of knowledge to narrowing the health and health service gap between culturally and linguistically diverse minority populations in the US. While there is a substantial amount data to indicate that health inequities exist and that they must be addressed, we need specific additional research data to guide an effective strategy to achieve health equity. Brown, Mohr, Gallo et al. (I 2013). JAIDS Computational Future of HIV Prevention in Minority Populations Perrino, T, Beardslee W, Bernal G, Brincks A, Cruden G, Howe G, Murry V, Pantin H, Prado G, Sandler I, Brown CH. Towards Scientific Equity for the Prevention of Depression and Internalizing Symptoms among Vulnerable Youth, Accepted for publication in Prevention Science 5

6 Five Key Principles and Concepts Collaborative Partnership Approach to Synthesis of Research Trials Collaborative Data Synthesis for Adolescent Depression Trials (CDSADT). Methods Development for Data Synthesis Integrative Data Analysis Treatment AND Prevention Impact on Populations Informing Science Policy : Scientific Equity Knowledge Exchange around Community, Practice, Policy, and Research Implementation Science 6

7 Our Goals of Data Synthesis for Randomized Trials Develop novel quantitative methods integrate findings across trials Apply these methods to prevent and treat depression in adolescence Building and testing the next generation of comprehensive intervention strategies. Is there Overall Benefit of One set of Interventions versus Control? Understand Who Benefits/Harmed in Which Contexts (Moderator), for How Long (Growth Modeling), How Benefit Occurs (Mediation) across What Outcomes? Which intervention(s) are better that others for which person (Comparative Effectiveness)? Brown et al. (2008). Methods for Testing Theory and Evaluating Impact in Randomized Field Trials: Intent-to-Treat Analyses for Integrating the Perspectives of Person, Place, and Time. Drug and Alcohol Dependence, S95: S74-S104,

8 How can we learn as much as we can from combining evidence from a family of similar randomized trials? Why the need for multiple trials? Single trials are almost always underpowered for interesting analyses. 4 times as many observations to achieve the same statistical power for a moderation analysis than it does for a main effect analysis. Why you don t need exact replication of interventions/dosage across trials? Combining even a small number of trials nearly always produces higher precision than a single trial - Brown et al., 2013 Prevention Science. Propose that we need Individual level data from multiple trials Ongoing Involvement of those who conducted the trials 8

9 Why Individual Level Data are Needed Three ways to conduct a data synthesis Meta-analysis Combine summaries of published statistical analyses from a family of randomized trials Requires retrieving reported summary analyses Integrative Data Analysis (IDA) -- Combine individual level data from a family of studies, e.g. trials Requires sharing of individual level data Parallel Data Analysis Have each lab conduct equivalent statistical analyses on their own dataset, then combine the summary findings into a composite analysis. Requires shared analytic method. Brown et al., 2013 Prevention Science 9

10 Meta-Analysis 10

11 Integrative Data Analysis 11

12 Parallel Data Analysis 12

13 Limitations in Meta-Analysis A. Theoretical results strongly favor integrative data analysis and parallel data analysis over meta-analysis for moderation and mediation. (Brown et al., 2013 Prev Sci, Dagne et al., under review) B. Empirical evidence also shows limitations in meta-analysis 13

14 Meta-Analyses and Overviews of Meta-Analyses Provide Little Information on Moderation, Mediation, and Comparative Effectiveness From Meta-Analytic Summaries From Original Trials Overview of 5 Meta-Analyses on Child Depression* Analyses Reported in Original Publications Types of Analysis ( Number ) Reported Reported / Possible Number (%) Significant Main Effects (35) 287/ (40.4%) Active vs. Control (31) 106 (46.1%) Comparative Effectiveness (4) 10 (17.5%) All Other Analyses (40) Subgroup (40) 108/ (36.1%) Moderation (0) 100/ (26.0%) Mediation (0) 21/ (66.7%) Total Distinct Analyses Reported (75) 516/ (60.3%) Sandler, I., Wolchik, S., Cruden, G., Mahrer, N., Brown, C.H., Brincks, A., Ahn, S. (2014). Overview of Meta-Analyses of the Prevention of Mental Health, Substance Use and Conduct Problems. Annual Review of Clinical Psychology, 10:

15 Meta-Analyses and Overviews of Meta-Analyses Provide Little Information on Moderation, Mediation, and Comparative Effectiveness From Meta-Analytic Summaries From Original Trials Overview of 5 Meta-Analyses on Child Depression* Analyses Reported in Original Publications Types of Analysis ( Number ) Reported Reported / Possible Number (%) Significant Main Effects (35) 287/ (40.4%) Active vs. Control (31) 106 (46.1%) Comparative Effectiveness (4) 10 (17.5%) All Other Analyses (40) Subgroup (40) 108/ (36.1%) Moderation (0) 100/ (26.0%) Mediation (0) 21/ (66.7%) Total Distinct Analyses Reported (75) 516/ (60.3%) Sandler, I., Wolchik, S., Cruden, G., Mahrer, N., Brown, C.H., Brincks, A., Ahn, S. (2014). Overview of Meta-Analyses of the Prevention of Mental Health, Substance Use and Conduct Problems. Annual Review of Clinical Psychology, 10:

16 Meta-Analyses and Overviews of Meta-Analyses Provide Little Information on Moderation, Mediation, and Comparative Effectiveness From Meta-Analytic Summaries From Original Trials Overview of 5 Meta-Analyses on Child Depression* Analyses Reported in Original Publications Types of Analysis ( Number ) Reported Reported / Possible Number (%) Significant Main Effects (35) 287/ (40.4%) Active vs. Control (31) 106 (46.1%) Comparative Effectiveness (4) 10 (17.5%) All Other Analyses (40) Subgroup (40) 108/ (36.1%) Moderation (0) 100/ (26.0%) Mediation (0) 21/ (66.7%) Total Distinct Analyses Reported (75) 516/ (60.3%) Sandler, I., Wolchik, S., Cruden, G., Mahrer, N., Brown, C.H., Brincks, A., Ahn, S. (2014). Overview of Meta-Analyses of the Prevention of Mental Health, Substance Use and Conduct Problems. Annual Review of Clinical Psychology, 10:

17 Can the potential benefits of Integrative Data Analysis be achieved? Are Trials Similar Enough to Synthesize? A. For Pharmacologic Treatment most trials have similar designs and interventions: Gibbons et al., 2012ab JAMA Psychiatry B. For Behavioral Interventions: All Prevention Trials and Majority of Treatment Trials, Challenging Methodologic Problems with Integrative Data Analysis Trials have different interventions: CB/IPT, Parenting, Child, Conjoint measures (harmonization) times of measurement (growth modeling) populations and contexts (comparability) Do trialists actually share their data? Pharmacologic Treatment: Gibbons et al., 2012ab JAMA Psychiatry Prevention: Prevention Science Special Issue Behavioral Prevention:

18 Do Antidepressants Work for Youth? In Single Trials, only one medication has demonstrated effectiveness in youth fluoxetine (Prozac) Meta-analysis of 27 pediatric trials (Bridge et al JAMA Psychiatry) On Depressive Symptoms 61% response to antidepressants vs 50% response to placebo (CI on gain 7%-15%) NNT = 10 18

19 Three Approaches to Synthesis of Trials beyond Meta- Analysis Repository of Trials Urge or require researchers to submit their trials so that others can obtain raw data Current NIMH Plan is to require all new trials funded by NIMH Access to All Trial Level Data through role as expert witness for US DoJ, Wyeth, Pfizer Pharmaceuticals (Gibbons) All Industry-Sponsored Fluoxetine trials in US Collaborative Synthesis Partnership of trialists around data harmonization and analysis Our main approach Received one youth Venlafaxine trial, not permitted to report impact 19

20 Summary of Youth Synthesis Findings of Fluoxetine on Depressive Sx Slope of Depression over 6 Weeks Reduced 16 units on placebo, 21.6 units on fluoxetine. Response Rates (50% Reduction in Symptoms) 6% for placebo, 30% for fluoxetine No indication of variation in impact on depressive symptoms as a function of baseline level of symptoms (CDRS-R) Gibbons RD, Hur K, Brown CH, Davis JM, & Mann JJ (2012)., JAMA Psychiatry. 20

21 NIMH funded Collaborative Data Synthesis on Adolescent Depression Trials Study Specific Aims: 1. Develop new statistical methods synthesize research findings on mediation (how) and moderation (for whom) across related randomized trials. 2. Apply these methods to identify shared and unique mediational mechanisms, and moderators of intervention effects on depression in adolescents. 3. Identify key gaps in our understanding of intervention efficacy and mechanism that need to be addressed in the next generation of intervention trials 21

22 Gathered Datasets 1. Prevention Synthesis Dataset has 19 prevention trials of 25 requested- approximately 76% shared 2. Treatment Synthesis Dataset has 14 treatment trials of 19 requested- approximately 74% shared This represents data from nearly 8,200 participants. Funding: National Institute of Mental Health R01-MH040859; Amy Goldstein- Project Officer 22

23 Gathered collaborative group of stakeholders in adolescent depression prevention and treatment Goals: 1. Sharing methods developments: Informing collaborators about synthesis methods issues being addressed; Making methods available to them 2. Working on substantive questions: Getting their input on prioritizing intervention mediator and moderators examined; Involving collaborators as manuscript co-authors; Disseminating findings to inform their work and trials. 3. Accurate representation of trials: Presenting trial level codings, summaries and hearing collaborators input 4. Recommendations for the next generation of trials: Work with collaborators to develop useful guidelines for future trials 23

24 Study Team Hendricks Brown (PI) George Howe (Co-I) Tatiana Perrino (Co-I) Hilda Pantin (Co-I) Ahnalee Brincks Gracelyn Cruden Getachew Dagne Peter De Chavez Ophelia Hernandez Shi Huang Lei Liu Juned Siddique Consultants William Beardslee David MacKinnon Irwin Sandler David Shern CDSADT Collaborators Funding: National Institute of Mental Health R01-MH040859; Amy Goldstein- Project Officer Robert Gibbons Scientific Advisory Board David Brent Guillermo Bernal Graham Emslie Velma Murry Other Partners Anne Sperling- NIMH Advocates David Shern, Mental Health America Individual Study Investigators William Beardslee Guillermo Bernal David Brent Steve Brunwasser Greg Clarke Bruce Compas Guy Diamond Tom Dishion Graham Emslie Judy Garber Jane Gillham Tracy Gladstone Nancy Gonzales John March Ann Mauricio Laura Mufson Velma Murry Hilda Pantin Guillermo Prado Cleve Redmond Irwin Sandler Richard Spoth Beth Stormshak Jose Szapocznik Jenn-Yun Tein Ben Van Voorhees Sharlene Wolchik Jamie Young 24

25 Collaborative Data Sharing Model Weekly research meetings Quarterly Collaborators Meeting Annual Face-to-Face Special Meetings Described in: Perrino, T., Howe, G., Sperling, A., Beardslee, W., Sandler, I., Shern, D., Pantin, H., Kaupert, S., Cano, N., Cruden, G., Bandiera, F., & Brown, C.H. (2013). Advancing Science through Collaborative Data Sharing and Synthesis. Perspectives on Psychological Science, 8(4): doi: / PMCID: PMC (page 437) 25

26 Key Elements Data Use Agreements, Human Subjects Protection Send/Receive the Data, Document, and Organize Understand trial intricacies Check Coding Decisions around Harmonization and Trial-Level Variables Prioritize synthesis work around critical questions Co-author papers 26

27 Why would Collaborators do all this for 5 years (and longer)? General Community-Based Participatory Research (CBPR) Trust, Respect, Equal Voice Mutual Self-Interest Saul Alinsky, Front seat to next generation trials Help their own analyses 27

28 Our Participatory Research Approach Differs from the Coming NIMH Policy on Depositing Trial Data in a Repository for others to analyze NIMH policy to make data widely available for others to analyze as quickly as possible We don t provide results for any single trial, not competing Imputed data on unmeasured outcomes provided back to individual investigators for their own analysis Coding decisions are continually reviewed for accuracy by trialists A great deal of Tacit Knowledge about these trials remains in the hands of the investigators 28

29 We all make mistakes in research Culumber ZW et al., Variation in Melanism and Female Preference in Proximate but Ecologically Distinct Environments. Ethology Volume 120, Issue 11, , November 2014 (Wiley Online Library) 29

30 Data Synthesis involving Individual Level Data Integrative Data Analysis 30

31 Consort Diagram on Prevention Trials: Inclusion/Exclusion: Population, Intervention, Trial Original trials requested (n= 24 ) Excluded (n= 6 ) Not meeting inclusion criteria (n= 0 ) Declined to participate (n= 5 ) Data not ready (n= 1 ) Additional trials requested to aid harmonization (n= 1) Original requests ( 18 ) 19 Included Trials Trials in dataset: 19 Participants in dataset: 5558 Excluded from analysis (n=0 ) Analysis Population: 11 17, non-clinical Intervention: Prevent internalizing or externalizing if family involved Design: Quality randomized trial with long term follow-up Excluded: Teacherdelivered intervention 31

32 Trial List Familias Unidas 1 (Hispanic) Familias Unidas II (Hispanic) Familias Unidas DJJ (Hispanic) Familias Unidas CDC (Hispanic) Family Talk Family Bereavement New Beginnings Program Bridges (Hispanic) Project Alliance I Project Alliance II K-IPT NARSAD IPT-AST ADEPT PODS Penn Resiliency Program I Penn Resiliency Program II CATCH-IT (Internet) Preparing for the Drug-Free Years 32

33 First Set of Research Questions on Prevention Trials 1. Are there intervention main effects of Prevention Programs on the adolescent depressive symptoms outcome? 2. Are there differential intervention benefits by adolescents gender age Race/Ethnicity baseline level of depressive symptoms 3. Are there differential intervention benefits by intervention: focus (i.e., depression-focus vs. non-depression focus as primary outcome) target (i.e., child alone, parent alone, child and parent together/conjointly) 33

34 Data 5,558 adolescents across 19 prevention trials w/ inclusion/exclusion criteria. 25 distinct, active intervention arms and 19 control conditions in these 19 trials. 33,045 assessments of depression over 19 trials through 6 years follo-up. 34

35 Methodologic Challenges Trials Use Different Measures to Assess Depressive Symptoms Trials Use Different Time Points to Assess Outcomes across Time 35

36 Depressive Symptom Measures by Trial Trial CDI Youth Self- Report Behavior Problem Checklist CESD CDRS PAL2 1 X 2 X 3 X 6 X X 8 X 10 X 12 X X 14 X X 17 X 18 X 28 X 49 X X 50 X X 61 X X 78 X X 84 X X 98 X 247 X X 698 X 36

37 Depressive Symptom Measures by Trial Trial CDI Youth Self- Report Behavior Problem Checklist CESD CDRS PAL2 1 X 2 X 3 X 6 X X 8 X 10 X 12 X X 14 X X 17 X 18 X 28 X 49 X X 50 X X 61 X X 78 X X 84 X X 98 X 247 X X 698 X 37

38 Participant Retention through 24 months 2.5 months 6.0 months months 16 months 24 months 38

39 Amount of Missing Data is Exceptionally Large 75% of the outcome measures (e.g., CDI) are missing across the 19 trials. 43% of the trials are missing ALL outcome measures across one of the 6 time windows. Missing by Design (Missing at Random) Approaches for this Paper Use a Single Factor Model for Internalizing/Depressive Symptoms Use Growth Curves to Fill in Across Time Full Information Maximum Likelihood to Handle Missing Variables 39

40 Measurement Model Across Time Testing for No Treatment Effect at Baseline CDRS CDI YSR Anxiety- Depression YSR Withdrawal- Depression CESD RPBC Anxiety- Withdrawal F0 Trial B = , p = Age Gender Race/Ethnicity Intervention Model Fit: LL = BIC = CFI =.964 SRMR =

41 Handling Time of Measurement Using Growth Modeling A few trials have very long follow-ups, only a few have data beyond 2 years of randomization so these times are dropped Trials have varying numbers and timing of assessments post randomization. 41

42 Transform Modeling Change through Growth Trajectories Need a Flexible Model Alternative Growth Patterns Slope Decreases over Time Slope Increases over Time Temporary Gain Offset by Worsening over Time Time 42

43 Transforming the Time Scale Empirically Determined Growth Model: Rapidly Increases then Slowly Increases, Gain is Permanent over 2 Years Intercept and Slope on this Transformed Scale 43

44 Longitudinal Intervention Model (Second-Order Growth Model) Measurement Model CDRS CDI YSR Anxiety- Depression YSR Withdrawal- Depression CESD RPBC Anxiety- Withdrawal Internalizing Time 0 Internalizing Time 2 Internalizing Time 3 Internalizing Time 4 Internalizing Time 5 Internalizing Time 6 Growth Model INTERCEPT SLOPE Covariates and Moderation Model Trial Age Male Race/Ethnicity Treatment 44

45 Summaries of Analysis Levels Individual Level Factors Time of measurement, which measures observed Age, Gender, Intervention Status Arm-Level Factors What type of intervention is delivered How it is delivered Trial-Level Factors Mean values for Intervention and Slope 45

46 Overall Treatment Effects: Significant Reduction in Symptoms over a Two-Year Period CDRS CDI YSR Anxiety- Depression YSR Withdrawal- Depression CESD RPBC Anxiety- Withdrawal Internalizing Time 0 Internalizing Time 2 Internalizing Time 3 Internalizing Time 4 Internalizing Time 5 Internalizing Time 6 INTERCEPT SLOPE 15% Faster Reduction Prevention Versus Control Trial Age Gender Treatment 46

47 Overall Effect: Interventions Reduce Symptoms Significantly Faster than does Control Condition Intervention Vs Control B = , SE = 0.178, p = 0.02 Difference in Slope* Between Intervention and Control Y = Slope intervention - Slope control *More negative values = desired effect 47

48 Moderated Effect: Depression Focuses Interventions Have Stronger Reduction in Symptoms than do Non-Depression Focused Interventions Intervention Difference in Slope* Between Intervention and Control B = (0.408), p = Y = Slope intervention - Slope control Depression- Focused 48

49 Latent Internalizing Latent Internalizing Depression vs. Non-Depression Intervention Focus Trajectory of Internalizing control intervention Trajectory of Internalizing by Intervention Target (0.593) p = Control: Nondepression target Tx: Non-depression target Control: Depression target Tx: Depression target -5 Time (log) -5 Time (log) 49

50 Intervention vs Control for Trials that are Depression Focused and Those that are Not Depression- Focused Intervention N= (0.593) p = Difference in Slope* Between Intervention and Control Y = Slope intervention - Slope control Non- Depression- Focused Intervention N=11 B = , SE = 0.340, p = Difference in Slope* Between Intervention and Control Y = Slope intervention - Slope control *More negative values = desired effect Hendricks excel 50

51 Internalizing Variation in Growth Trajectories and Differential Impact on Initially Higher Internalizing and Lower Internalizing Latent Class Trajectories Control 37% 4 Intervention Time Control Intervention 63% 51

52 Intervention Impact on African American and Hispanic Youth 52

53 Intervention Effect on African Americans and for Hispanics/Latinos in trials not designed specifically for Hispanics Overall African Americans (n = 491) Hispanics/Latinos (n = 286) Effects and Confidence Bounds Select Race and Ethnic Group Overall African American Hispanic Slope of Symptom Change 53

54 Impact of a Hispanic-Focused Interventions on Hispanic Youth: 3 Familias Unidas Trials Demonstrate Significant Benefit on Depressive Sx When Hypothesized Mediator of Family Communication is Poor Significant Sx Reduction and Mediation Through Family Communication Perrino, Pantin et al Prevention Science No Sx Reduction and No Mediation Through Family Communication 54

55 There are few trials specifically designed for African Americans or Hispanics Health (and Health Service ) Disparities Health Equity Scientific Equity 55

56 What Evidence is Available to Show Whether Prevention Programs/Strategies Work for Hispanics? Race/Ethnic Group In 19 Preventive Trials In NIMH, NIDA, NIAAA Funded Trials Specific to this Population Percent In US Population of Youth African American 491 (9%) 8 (4%) 17% Hispanic 286 (5%) general (25%) Hispanic specific 16 (9%) 21% Other 4,781 (61%) 159 (87%) 62% Total 5, % 56

57 Child-Focused Preventive Interventions Have Stronger Effects on Depressive Sx Compared to Non-Child Focused Intervention Vs Control B = , SE = 0.412, p = 0.04 Child Focused vs Non-Child Focused Difference in Slope* Between Intervention and Control Y = Slope intervention - Slope control *More negative values = desired effect 57

58 Summary & Conclusions Collaborative Data Synthesis is Viable and Successful Methodology Partnership that redefines relationship between researchers, policy makers, advocates, practitioners Early Findings Prevention Effective and Sustained Over 2 Years. Significant Impact for both lower and higher depressive symptoms Interventions deliberately focusing on depression have much stronger impact Important moderators and mediators Child focused interventions stronger overall effect Family-based interventions not effective for everyone BUT They are successful when parent-child communication is poor Policy Implications Prevention, Integrated with Treatment for Population-Level Impact Multiple prevention interventions to match youth and family needs Scientific Equity is far from being Achieved Hispanic-Focuses Interventions targeting Poor Family Communication Improve Internalizing and Externalizing No Clear Evidence that African-American or Hispanic Youth Gain from Interventions Not Specific to these Populations 58

59 To Address Policy Needs, it Requires an Integration/Alignment of Research, Practice, Policy, Advocacy, and Methodology Policy Determine Priorities Funding Coordination Research Conduct Research on Implementation Home of Prevention Community Resources Practice Synthesis Quality Improvement Implementation Political Will and Strategy Methodology Community and Advocacy 59

60 CDSADT Publications Brown, C. H., Kellam, Sheppard G., Kaupert, Sheila, Muthen, Bengt, Wang, Wei, Muthen, Linda, Chamberlain, Patricia, PoVey, Craig, Cady, Rick, Valente, Thomas, Ogihara, Mitsunori, Prado, Guillermo, Pantin, Hilda, Szapocznik, Jose, Czaja, Sara, & McManus, John. (2012). Partnerships for the Design, Conduct, and Analysis of Effectiveness, and Implementation Research: Experiences of the Prevention Science and Methodology Group. Administration and Policy in Mental Health and Mental Health Services Research, 39, 4: PMID PMCID: PMC Brown. C.H., Sloboda, Z., Faggiano, F., Teasdale, B., Keller, F., Burkhart, G., Vigna-Taglianti, F., Howe, G., Masyn,K., Wang, W., Muthén, B., Stephens, P., Grey, S., Perrino, T., and the Prevention Science and Methodology Group. (2013). Methods for Synthesizing Findings on Moderation Effects across Multiple Randomized Trials. Prevention Science, 14(2): doi: /s PMID PMCID: PMC Perrino, T., Beardslee, W.R., Bernal, G., Brincks, A., Cruden, G., Howe, G., Murry, V., Pantin, H., Prado, G., Sandler, I., Brown, C H. (2014) Toward Scientific Equity for the Prevention of Depression and Depressive Symptoms in Vulnerable Youth. Prevention Science, Online First 28 October Perrino, T., Howe, G., Sperling, A., Beardslee, W., Sandler, I., Shern, D., Pantin, H., Kaupert, S., Cano, N., Cruden, G., Bandiera, F., & Brown, C.H. (2013). Advancing Science through Collaborative Data Sharing and Synthesis. Perspectives on Psychological Science, 8(4): doi: / PMCID: PMC Perrino, T., Pantin, H., Prado, G., Huang, S., Brincks, A., Howe, G., Beardslee, W., Sandler, I., Brown, C.H. (2014). Preventing Internalizing Symptoms Among Hispanic Adolescents: A Synthesis Across Familias Unidas Trials. Prevention Science. doi: /s (In Processing in NIHMS, Waiting for release to PMC) NIHMS ID: Perrino, T, Beardslee W, Bernal G, Brincks A, Cruden G, Howe G, Murry V, Pantin H, Prado G, Sandler I, Brown CH. (Accepted for Publication). Towards Scientific Equity for the Prevention of Depression and Internalizing Symptoms among Vulnerable Youth. To appear in Prevention Science. Sandler, I., Wolchik, S., Cruden, G., Mahrer, N., Brown, C.H., Brincks, A., Ahn, S. (2014). Overview of Meta-Analyses of the Prevention of Mental Health, Substance Use and Conduct Problems. Annual Review of Clinical Psychology, 10 doi /annurev-clinpsy PMID: PMCID: PMC Siddique, J., Chung, J.Y., Brown, C. H., & Miranda, J. (2012). Comparative effectiveness of medication versus cognitive-behavioral therapy in a randomized controlled trial of low-income young minority women with depression. Journal of Consulting and Clinical Psychology, 80(6): PMID PMCID Siddique, J., Harel, O. & Crespi, C.M. (2013). Addressing Missing Data Mechanism uncertainty Using Multiple-Model Multiple Imputation: Application to a Longitudinal Clinical Trial. Annals of Applied Statistics. PMID: PMCID: PMC Siddique, J., Harel, O., Crespi, C.M., Hedeker, D. (2012) Binary variable multiple-model multiple imputation to address missing data mechanism uncertainty: Application to a smoking cessation Trial. Statistics in Medicine. 6(4): PMID: NIHMSID: (Release to PMC 12 months after publication) 60

61 PSMG Publications Cross, W., West, J. Wyman, PA, Schmeelk-Cone, K., Xia, Y., Tu, X., Teisl, M., Brown, CH., & Forgatch, M. (Accepted). Observational measures of Implementer Fidelity for a School-based Preventive Intervention: Development, Reliability and Validity. Prevention Science. NIHMS ID: , Journal Submitting. Dagne, G. (2013). Bayesian Inference for Skew-Normal Mixture Models with Left-Censoring. Journal of Biopharmaceutical Statistics, 23(5): PMID: PMCID: PMC Dagne, G. & Huang, Y. (2013). Bayesian Semiparametric measure Tobit Models with Left-Censoring, Skewness, and Covariate Measurement Errors, Statistics in Medicine, 32(22): epub online 2 April PMID: PMCID: PMC Flory, K., Malone, P.S. & Lamis, D. (2011). Childhood ADHD Symptoms and Risk for Cigarette Smoking during Adolescence: School Adjustment as a Potential Mediator, Psychology of Addictive Behaviors, 25(2): PMID: PMCID: PMC Gibbons, RD, Brown, CH, Hur, K, Davis, JM, & Mann, JJ (2012). Suicidal Thoughts and Behavior with Antidepressant Treatment: Reanalysis of the Randomized Placebo-Controlled Studies of Fluoxetine and Venlafaxine. Archives of General Psychiatry (69)6: doi: /archgenpsychiatry PMID PMCID: PMC Gibbons RD, Hur K, Brown CH, Davis JM, & Mann JJ (2012). Who Benefits from Antidepressants?: Synthesis of 6-Week Patient-Level Outcomes from Double-Blind Placebo Controlled Randomized Trials of Fluoxetine and Venlafaxine. Archives of General Psychiatry, 69(6): doi: /archgenpsychiatry PMID PMCID: PMC Gibbons, R.D., Hur, K., Brown, C. Hendricks, Mann, JJ. (2009). Relationship Between Antiepileptic Drugs and Suicide Attempts in Patients With Bipolar Disorder. Archives of General Psychiatry, 66(12): doi: /archgenpsychiatry PMID: PMCID: PMC Gibbons, R.D., Hur, K., Brown, C. Hendricks, Mann, JJ. (2010). Antiepileptic drugs and suicide attempts in patients with bipolar disorder-reply. Archives of General Psychiatry, 67(12): PMID: PMCID: PMC Huang, Y., & Dagne, G. (2012). Bayesian semiparametric nonlinear mixed-effects joint models for data with skewness, missing responses, and measurement errors in covariates. Biometrics. 68(3): doi: /j x. PMID PMCID: PMC Huang, Y., & Dagne, G. (2011). Simultaneous Bayesian inference for skew-normal semiparametric nonlinear mixed-effects models with covariate measurement errors. Bayesian Analysis, 6 (4), PMID PMCID: PMC Imai, K., Jo, B., & Stuart, E. A. (2011). Commentary: Using potential outcomes to understand causal mediation analysis. Multivariate Behavioral Research 46(5): Jo, B. (in press). Growth mixture modeling and causal inference. Forthcoming in G. R. Hancock and J. R. Harring (Eds.), Advances in Longitudinal Methods in the Social and Behavioral Sciences. Information Age Publishing. Jo, B., & Stuart, E. A. (2012). Causal interpretations of mediation effects: Comment on "Principal stratification as a framework for investigating mediational processes in experimental settings" by Lindsay C. Page. Journal of Research on Educational Effectiveness, 5(3): Jo, B., Stuart, E. A., Mackinnon, D., & Vinokur, A. (2011). The use of propensity scores in mediation analysis. Multivariate Behavioral Research, 46, PMID PMCID: PMC Jo, B., & Vinokur, A. (2011). Sensitivity analysis and bounding of causal effects with alternative Identifying assumptions. Journal of Educational and Behavioral Statistics, 36, PMID PMCID: PMC Masyn, K. (2009). Discrete-Time Survival Factor Mixture Analysis for Low-Frequency Recurrent Event Histories, Research in Human Development, 6: NIHMS (Awaiting author approval in NIHMS) Masyn, K., Henderson, C.E. & Greenbaum, P.E. (2010). Exploring the Latent Structures of sychological Constructs in Social Development Using the Dimensional-Categorical Spectrum, Social Development, 19(3): NIHMS Stuart, E. A. & Jo, B. (2011). Assessing the sensitivity of methods for estimating principal causal effects. Statistical Methods in Medical Research, epub ahead of print 3 October PMID: PMCID: PMC Stuart, E.A. & Ialongo, N.S. (2010). Matching methods for selection of subjects for follow-up, Multivariate Behavior Research, 45: PMID: PMCID: PMC Wang, C., Jo, B., Brown, C.H. (2014). Causal Inference in Longitudinal Comparative Effectiveness Studies with Repeated Measures of A Continuous Intermediate Variable. Statistics in Medicine. Epub ahead of print 2014 Feb 27. doi: /sim.6120 PMID: NIHMS ID: In Process, Journal Submitting. 61