1. Clinical Trials for Comparative Effectiveness Research Mark Hlatky MD Mark Hlatky MD Stanford University January 10, 2012

2. What is CER?  Generation and synthesis of evidence that  compares the benefits and harms  of alternative methods to prevent, diagnose, treat, and monitor a clinical condition  or to improve the delivery of care.  The purpose of CER is to  assist consumers, clinicians, purchasers, and policy makers  make informed decisions  that will improve health care  at both the individual and population levels. Institute of Medicine, 2009

3. CER Methods  CER can use very different research design  Randomized trials  Observational studies using registries, claims, or electronic health records  Simulations, models, meta-analysis  Randomized trials accepted as best approach because of control of confounding

4. RCT Limitations  Most trials due for regulatory approvals, and not good for CER  Use placebo, not “next best therapy”  Use ideal patients, not typical ones  Lab measures as endpoints  Safety poorly assessed

5. CER Trials  Limitations of RCTs for CER are not inherent  Can compare new therapy with an active control  Can enroll representative patients  Can focus on patient-centered outcomes

6. Representative trials  Generalizability of RCT results is always controversial  Inclusion/exclusion criteria can be minimized to broaden patient selection  Also important to include representative practice sites, particularly for trials of procedures  Volume-outcome relationship  Centers of excellence vs typical practice locations

7. Large Practical Trials  Larger sample sizes needed for CER trials  Use of active controls narrows expected superiority margin  “Equivalence trials ”need large sample sizes to narrow confidence intervals  Analysis of subgroups requires more patients  Trials need to be simpler and less expensive to be practical for CER

8. Reducing Trial Costs  Data collection is expensive, so minimize trial-specific data  Compact, simple data forms  Rely on existing mechanisms to capture data at baseline and during follow-up (hybrid trials)  Clinical registries (esp baseline data)  Electronic health records  Claims data

9. Clinical Registries  Registries are a middle ground between RCTs and regular care  Data elements standardized, collected as part of routine care  e.g., ACC PCI registry, STS CABG registry  Could use registry to collect clinical data, and just add randomization

10. Follow-Up Data  Registries typically collect data from acute care or procedures  Long-term follow-up needed  Link to claims, EHRs, or mortality files  Is elaborate adjudication of non-fatal outcomes necessary, or could claims be used instead?  Web follow-up for symptoms, functional status?

11. Cluster RCTs  CER also interested in quality of care, health delivery mechanisms, etc.  Can randomize practice sites to alternative care approaches (e.g., disease management)  Could also use cluster trials for “strategy trials” or default management approaches to common conditions (eg diabetes, hypertension)

12. Practical, Cluster RCT: Systems Approach to Heart Failure Care

13. Conclusions  RCT is one of several designs for CER studies  Larger trials generally needed for CER  Newer designs for RCTs needed to keep costs down and make CER feasible