1. Clinical Trials: Introduction from an Epidemiologic Study Design Perspective Health Sciences Center Health Sciences Center School of Public Health & Stanley S. Scott Cancer Center LSU Neal Simonsen Neal Simonsen

2. Experimental Study Design Exposed Not exposed Disease nonoccurrence Unethical to perform experiments on people if exposure is harmful Exposure assigned http://www.epiet.org/course/Presentations/Presentations%202003/13-%20Cohort&Case%20control/13-%20Cohort&Case%20control_files/frame.htm OK Not OK

3. A more-often-ethical alternative for conducting studies in human populations Observational Epidemiology allows capitalization on “natural” or “unplanned” experiments.   Take advantage of groups who have been exposed for non-study purposes.

4. What is a cohort? Group of individuals Group of individuals  …sharing same experience  …followed up for specified period of time Examples: Examples:  Birth cohort  Cohort of guests at barbecue  Occupational cohort of electricians http://www.epiet.org/course/Presentations/Presentations%202003/13-%20Cohort&Case%20control/13-%20Cohort&Case%20control_files/frame.htm

5. Cohort Study Exposed Not Exposed Develop Disease Do Not Develop Disease Develop Disease Do Not Develop Disease

6. First select Exposedaba + b a + b c Not exposedcdc + d c + d Then follow to see whether… Disease Disease does developsnot develop a Totals Incidence Rates of Disease Analytical Design of a Cohort Study

7. Relative Risk Risk in exposed Risk in non-exposed

8. Potential Biases in Cohort Studies Bias in the assessment of the outcome Bias in the assessment of the outcome  If person assessing disease also knows exposure o May be addressed by blinding Information bias Information bias  Quality of information not comparable between exposed and non-exposed individuals

9. Biases in Cohort Studies (Cont’d) Biases from non-response and losses to follow-up Biases from non-response and losses to follow-up  Non-participation / non-response and dropouts can complicate interpretations of findings Analytic bias Analytic bias  Biases of epidemiologists/biostatisticians analyzing the data

10. Relative merits: cohort studies Advantages Clear temporal relationship Clear temporal relationship Least susceptible to some forms of bias Least susceptible to some forms of bias Can examine multiple predictors of outcome Can examine multiple predictors of outcome Efficient for rare exposures Efficient for rare exposures Useful when RCT infeasible, unethical Useful when RCT infeasible, unethical Disadvantages No control over exposure (vs. RCT) Inefficient for rare or long-latent diseases Loss to follow-up threatens validity Potential bias in outcome ascertainment Relatively resource- and time-intensive http://www.healthsystem.virginia.edu/internet/MTPCI/Introcourse02/Lecture%206%20Bovbjerg.ppt

11. Basic cohort design Crucial comparison: outcomes in persons with vs. without predictor variable (or with different levels of predictor) Study population Predictor present Predictor absent No Yes No Yes follow-up period outcome of interest no investigator control http://www.healthsystem.virginia.edu/internet/MTPCI/Introcourse02/Lecture%206%20Bovbjerg.ppt

12. Basic trial design Estimate of effect is rate (risk) of outcome in treatment vs. control (e.g. risk[treatment]/risk[control]) Study population Treatment No treatment (usual care, placebo) No Yes No Yes follow-up period outcome of interest assignment by investigator http://www.healthsystem.virginia.edu/internet/MTPCI/Introcourse02/Lecture%206%20Bovbjerg.ppt

13. Two trial flavors Trials may be: Trials may be:  Non-randomized (in terms of assignment to treatment group) or  Randomized Why be random? Why be random? o [Sounds sooo unscientific…]

14. Example of Bias due to Non- randomized Treatment Assignment Study in Lanarkshire, 1930 Study in Lanarkshire, 1930  Treatment 1: 3/4 pint raw milk/day (n=5,000)  Treatment 2: 3/4 pint pasteurized milk/ day (n=5,000)  Treatment 3: no milk (n=10,000)  Response = growth (increase in height and weight)

15. The Results  No milk group had significantly better growth than either milk treated group  Giving kids milk is bad for their growth???

16. The Problem and its Consequences  Assignment initially randomized, but… teachers could “improve” the assignment  Results: no milk group had better growth than either milk treated group  Why? öTeachers tended to assign poorer children to the milk treated groups

17. Types of Associations Milk supplementation A. CausalB. Due to Confounding Real vs. Spurious Associations Milk supplementation Better Growth Poorer Growth Poverty Observed Association Observed Association

18.  “Gold Standard”: Basic randomized, controlled clinical trial design Estimate of effect is rate (risk) of outcome in treatment vs. control (e.g. risk[treatment]/risk[control]) Study population Treatment No treatment (usual care, placebo) No Yes No Yes follow-up period outcome of interest random assignment by investigator http://www.healthsystem.virginia.edu/internet/MTPCI/Introcourse02/Lecture%206%20Bovbjerg.ppt

19. Summary: Epidemiologic Study Types Studies of group characteristics Studies of group characteristics  Ecologic  Potential ecologic fallacy Studies of individual characteristics* Studies of individual characteristics*  Cross-sectional  Case-Control  Cohort  Intervention/Clinical Trial * [~within one group vs. another(s)]