Presentation Transcript

1. Overview of Study Designs

2. Study Designs Experimental Observational Analytical Group Randomized Trial Descriptive Randomized Controlled Trial Cohort Case-control Cross-sectional Ecological

3. Experimental v. Observational • “Gold Standard” • Investigator controls exposure • Less common • Investigator observes exposures • Most common

4. Experimental Study: Randomized Controlled Trial INTERVENTION GROUP OUTCOME ASSESSMENT Randomly Assigns Intervention Comparison Future Present TIME

5. Randomized Controlled Trial • Randomization is random distribution or allocation of subjects into groups • Treatment or intervention group • Comparison group (placebo, another tx, no tx, usual care) • Helps ensure that both groups are similar re: known and unknown factors • Two types of RCTs • Preventive trials • Therapeutic trials

6. Group Randomized Trial • Unit of assignment = group • Unit of analysis = individual (usually) • Investigator controls exposure status • Randomly assigns group to intervention group • Exposure assessed before outcome

7. Strengths Able to demonstrate causal association Randomization deals with known and unknown confounders Investigators directly control intervention Limitations Ethical concerns (equipoise) Gen practice ≠ artificial Expensive Power often an issue Experimental Study

8. Observational Studies • Most common, by necessity • Observes nature without intervention • Takes advantage of people’s natural exposures (choice, occupation, environment, residence) • Collect data and statistically analyze results • Goal is to “mimic” experimental study • Statistics more important and usually more complex • Can be descriptive or analytic

9. Study Designs Experimental Observational Analytical Group Randomized Trial Descriptive Randomized Controlled Trial Cohort Case-control Cross-sectional Ecological

10. Cohort Studies • Two ways to assemble cohort: • Group of individuals with common characteristic or experience (then assess exposure before outcome) • Select subjects without outcome according to exposure status (esp important for rare exposure) • Followed over time to determine incidence of symptoms, disease, or death • Analysis focuses on risk of outcome in exposed compared to unexposed groups • Unexposed is referent or comparison group

11. Prospective Cohort Studies • Only include outcome-free individuals • At risk of developing outcome • Followed into future to observe outcomes • Can take long time to complete • Important to re-assess exposures over time • Useful design for rare exposures • Not for rare outcomes • Advantage: exposure assessed before subject knows outcome status; temporal sequence

12. Prospective Cohort Study EXPOSURE ASSESSMENT OUTCOME ASSESSMENT Exposed Unexposed Future Present TIME

13. Retrospective Cohort Studies • Both exposure and outcome occur before study • Studies only prior outcomes • Historical cohort reconstructed from existing data sources before study begins • Example: Effects of pesticide exposure on cancer mortality (cohort of factory workers employed by manufacturer identified from personnel records) • Need good records with info on many variables • Sometimes good outcome info not available

14. Retrospective Cohort Study OUTCOME ASSESSMENT EXPOSURE ASSESSMENT Exposed Unexposed Present Past TIME

15. Strengths Easier to differentiate cause from effect Direct estimation of incidence Able to examine multiple outcomes Efficient for rare exposures Prospective Reduces recall bias Limitations Large sample size Prospective Long-term follow-up Expensive Cohort Study

16. Case-Control Studies • Both exposure and outcome have occurred • Classified according to outcome status before exposure ascertained • Cases with outcome selected from well-defined source population • Controls without outcome* sampled from population that produced cases • Analysis focuses on odds of exposure in cases compared to controls

17. Case-Control Study • Grouped on basis of past or current outcome • Outcome already occurred before exposure assessed • May lead to biased recall • May be difficult to establish temporality • Useful design for rare outcomes • Not for rare exposures • Advantages: • Less expensive than cohort (usually) • Can look at multiple exposures

18. Case-Control Study EXPOSURE ASSESSMENT OUTCOME ASSESSMENT Cases (with outcome) Controls (without outcome)* Past Present TIME

19. Strengths Efficient for rare outcomes Able to examine multiple exposures Limitations ↑ possibility of bias Temporal association Inefficient for rare exposures Case-Control Study

20. Cross-Sectional Studies • Study population not selected based on outcome or exposure status • Outcome and exposure assessed at same time • Snapshot at single point in time • Problems inferring temporal sequence • Identify prevalent cases of long duration

21. Analytical Cross-Sectional Study OUTCOME ASSESSMENT EXPOSURE ASSESSMENT Exposed Unexposed Past Present TIME

22. Strengths Less time and resources No follow-up time involved More representative of well-defined general population Useful when onset is difficult to establish (incidence) Limitations Temporal association Potential for prevalence-incidence bias Not for rare exposures or outcomes Reverse causality Cross-Sectional Study

23. Analytical Ecological Study OUTCOME ASSESSMENT EXPOSURE ASSESSMENT Exposed Unexposed Past Present TIME Note: Often in ecological studies, the exposures and outcomes are continuous measures

24. Strengths Relatively quick and inexpensive May be only appropriate design for research question May be useful when intra-group variability in exposure is small May provide greater inter-group variability across larger ecological units Limitations Ecological fallacy Secondary data sources Confounding Temporal association Migration across groups Ecological Study

25. Characteristics of Study Designs • Exposure(s) of interest? • Outcome(s) of interest? • Population investigated? • Recruitment? • Study design? • Strengths? • Limitations?