1. 1 Diabetes Multi-Center Research Consortium (DMCRC)  Coordinating Center  HMO Research Network DEcIDE Center  PI Joe Selby, MD  Co-PI Patrick O’Connor MD  Affiliate Center  Johns Hopkins University DEcIDE Center  PI Jodi Segal, MD  Co-PI Eric Bass, MD

2. 2 The Case for CER in Diabetes The Case for CER in Diabetes HIGH BURDEN OF DISEASE HIGH BURDEN OF DISEASE – High, rising prevalence of diabetes (>23 million diagnosed cases, 10% prevalence in adults) – Chronicity – life expectancy with diabetes >20 years; age at diagnosis decreasing; complication-related morbidities lead to many years with high annual costs

3. 3 The Case for CER in Diabetes UNCERTAINTY  Variation in Practice UNCERTAINTY  Variation in Practice – Multiple therapeutic choices (6 classes of oral agents, two classes of injectables) – Several options are relatively new and costly – Treatments vary in mechanisms of action, relative effectiveness and safety uncertain – Optimal treatment “strategies” unclear: timing of pharmacotherapy; treatment targets; sequencing and combination TX

4. 4 The Case for CER in Diabetes Complexity in Optimizing Effectiveness Complexity in Optimizing Effectiveness – Self-care, including medication adherence is central to effectiveness, but difficult to optimize – Out-of-pocket medication costs interfere with medication adherence and self-care – Blood pressure, lipid control, and aspirin each more effective than “tight” glycemic control in preventing most diabetic complications – Weight management is important, but several medication classes cause weight gain

5. 5 The Case for CER in Diabetes Complexity in Optimizing Effectiveness Complexity in Optimizing Effectiveness – “Systems Approaches” may enhance self-care and improve adherence and care coordination – Depression common in diabetes, but role of depression therapy in improving control unclear – Role of “tight” control in preventing CVD complications thrown into question in 2008 by three RCT’s: ACCORD, ADVANCE, VADT – Other adverse consequences of tight control - wt. gain, hypoglycemia, fractures – Benefits may vary by patient age, DM duration

6. 6 The Case for CER in Diabetes PREVENTION AND EARLY DETECTION PREVENTION AND EARLY DETECTION – Reservoir of undiagnosed cases, but the net benefits of screening various populations for diabetes not entirely clear – Diabetes can be prevented or postponed by lifestyle and/or pharmacotherapy; but optimal “real world” programs not fully clarified

7. 7 Data Committee Clinical Committee Methods Committee Stakeholder Committee Administrative Committee Project Manger Executive Committee – Includes AHRQ, Coordinating, Affiliate Center Leadership DMCRC Structure

8. 8 Data Committee Clinical Committee Methods Committee Stakeholder Committee Administrative Committee Project Manger Executive Committee – Includes AHRQ, Coordinating, Affiliate Center Leadership DMCRC Structure

9. 9 Expanded Executive Committee Also includes: Also includes:  Vanderbilt DEcIDE Center – Marie Griffin MD, PI – Comparative Effectiveness of Oral Agents in Type 2 Diabetes  RTI DEcIDE Center – Suzanne West Ph.D. – Comparative Effectiveness of Oral Hypoglycemics on Chronic Kidney Disease and on Time to Initiation of Maintenance Insulin

10. 10 DMCRC Work Assignments 1. Comparative Effectiveness of Bariatric Surgery vs. Usual Care in Type 2 Diabetes (two projects) 2. Proposal for New Statistical Briefs - using representative data to characterize trends in diabetes treatment and outcomes (joint) 3. Form and Convene Stakeholders’ Group (HMORN) 4. Form and Convene Data Committee (JHU) – with HMORN, Vanderbilt, RTI participation 5. Comparative Effectiveness Study of Intensive Glycemic Control vs. Less Intensive Control in presence vs. absence of tight blood pressure and lipid control (two projects)

11. 11 DMCRC Stakeholder Committee Government Agencies – AHRQ, NIDDK, CMS, FDA, CDC, VA Government Agencies – AHRQ, NIDDK, CMS, FDA, CDC, VA Clinicians – ACP,AAFP, AADE Clinicians – ACP,AAFP, AADE Patients - ADA, individual patient rep. Patients - ADA, individual patient rep. Expanded DMCRC Executive Committee Expanded DMCRC Executive Committee

12. 12 Stakeholder - Developed Priorities 1.Role of intensive glucose control in individuals with and without CVD, not typically enrolled in trials 2.Comparative effectiveness of multi-risk factor reduction on long-term CV outcomes 3. Comparison of system-based (coordinated) care vs. usual care 4.Approaches to DX and treatment of depression in diabetes 5.Risk factors for nonadherence – effects of nonadherence on costs and clinical outcomes

13. 13 Stakeholder - Developed Priorities 6. Effectiveness of eliminating co-pay for effective drugs (statins, ACE-I’s, beta blockers, anti-diabetic meds) – on outcomes and total drug burden? 7. Patient reported outcomes, HRQoL in relation to therapy 8. Optimal timing for metformin initiation on the continuum of pre-DM -> DM 9. Best strategies for behavior change. Who should do it and where should it be done? 10. Understanding patient attitudes toward insulin use

14. 14 Work Assignment #1: Health outcomes of bariatric surgery in individuals with type 2 diabetes HMORN: PI: David Arterburn MD (Group Health Cooperative) (Group Health Cooperative) Johns Hopkins U: PI: Jodi Segal MD

15. 15 WA #1: Primary Aims Compare short-term outcomes between patients under- going BS and comparable patients who don’t Compare short-term outcomes between patients under- going BS and comparable patients who don’t – Resolution of diabetes (no meds, nl FPG’s – Medication use – BMI Change – Glycemic, BP, and lipid Control Compare longer-term outcomes between patients under- going BS and comparable patients who don’t: Compare longer-term outcomes between patients under- going BS and comparable patients who don’t: – Recurrence of diabetes (abnormal labs or re-initiation of diabetes medications) – Death, hospitalization, re-operation Examine differences in these outcomes by type of BS: Bypass, banding, gastric sleeve Examine differences in these outcomes by type of BS: Bypass, banding, gastric sleeve

16. 16 WA #1: Secondary Aims Compare a variety of shorter- and longer-term outcomes between patients under- going BS and comparable patients who don’t (HMORN and JHU): Compare a variety of shorter- and longer-term outcomes between patients under- going BS and comparable patients who don’t (HMORN and JHU): – Development and progression of CKD and DN – Development and progression of diabetic retinopathy – Development of incident cardiovascular disease – Long-term health care utilization – Incidence of various cancers – Incidence of osteoporotic fracture – Incidence of urolithiasis Examine differences in these outcomes by type of BS: Bypass, banding, gastric sleeve Examine differences in these outcomes by type of BS: Bypass, banding, gastric sleeve

17. 17 WA #1: Study Design: BSMed Mean Age (yrs) 5447 % Female 5680 Median BMI 4638  Cohort Study in 180,000 patients with evidence of Type 2 diabetes, BMI >35, aged 18-30  Note: presence of BMI in EMR required  Approximately 3,100 BS with BMI 2002 – 08

18. 18 WA #1: The Cohort Enters cohort when T2 DM and BMI > 35 identified Bypass Banding Sleeve No BS 2002-2008 BS No BS End 2009

19. 19 WA #1: Analysis Plan Propensity Score (time dependent) calculated for each cohort member Propensity Score (time dependent) calculated for each cohort member Probabilities associated with each decile of PS examined, with possible trimming of very low probability deciles Probabilities associated with each decile of PS examined, with possible trimming of very low probability deciles Modeling of outcomes in remaining cohort examined using time-varying predictors for BS and key covariates Modeling of outcomes in remaining cohort examined using time-varying predictors for BS and key covariates For comparisons by type of surgery, separate cohort analyses restricted to persons having BS For comparisons by type of surgery, separate cohort analyses restricted to persons having BS Treatment heterogeneity examined by age group, presence of prior comorbid conditions Treatment heterogeneity examined by age group, presence of prior comorbid conditions

20. 20 WA #1: Key Points in Analysis Multi-variable models predicting outcome will NOT use PS Multi-variable models predicting outcome will NOT use PS For discrete analyses, models will evaluate non- proportional (i.e., time-varying hazards) For discrete analyses, models will evaluate non- proportional (i.e., time-varying hazards) Will also examine effect heterogeneity by year of surgery and volume of surgeon Will also examine effect heterogeneity by year of surgery and volume of surgeon Many more BS patients without pre-surgical BMI, who may contribute to some analyses where BMI less likely to confound. Many more BS patients without pre-surgical BMI, who may contribute to some analyses where BMI less likely to confound.