Study designs in epidemiology Focus on Randomized Controlled Trials (RCT)

1. Study designs in epidemiologyFocus on Randomized Controlled Trials (RCT) • Marike van der Leeden, PhD

2. The study question decides what design to choose The research question

3. What are you planning to do? PICO P - who are the patients or what’s the problem? I - what is the intervention or exposure? C – what is the comparison group? O - what is the outcome or endpoint?

4. Example Is there a difference in the efficacy between physiotherapy and manual therapy in patients with non-specific lowback pain?

5. Key questions in epidemiological studies • How many are (becoming) ill? (occurence) • Why do some people become ill why others stay healthy? (etiology/causality) • How can we determine if somebody has a specific health condition? (diagnostics) • What can we do to improve the health condition of individuals/populations? (effect) • What will happen to those who are ill? (prognosis)

6. Observational studies

7. Cross-Sectional Design Takes a measure of a population at a certain point in time (e.g., a survey research) O1 Are easy, fast, and inexpensive Descriptive, eg can determine prevalence of disease Can look at associations between variables Cannot determine cause and effect Quantitative Research 6

8. Longitudinal studies • Repeated measures over time • O1 O2 O3 ….. • Data collection timing: prospectively, retrospectively, combination • Cohort study: determinant = starting point leading to disease or disease status • Case-control study: disease or disease status = starting point

9. Cohort Studies Data obtained from groups who have already been exposed, or not exposed, to the factor of interest. No allocation of exposure is made by the researcher. Best for studying effects of risk factors on an outcome/looking for causes of disease/health outcomes • Merits: several outcomes could be studied at the same time • Limitations (of prospective type): expensive; time-consuming; inefficient for rare diseases

10. Cohort Design disease Factor present no disease Study population free of disease disease Factor absent no disease present future time Study begins here

11. Minimising bias and confounding Bias - caused by systematic variation/error in selecting patients, measuring outcomes, analysing data – take extra care Confounding - factors that affect the interpretation of outcomes eg people who carry matches are more likely to develop lung cancer, but smoking is the confounding factor – so measure likely confounders too

12. The beast of bias – Andy Field • https://www.youtube.com/watch?v=GKmvz6SbBoo&index=3&list=PL343F1B5F55734D55

13. Experimental studies • Studying the impact of an intervention eg treatment or health service • Is intervention (therapy, preventive action, supply) effective? –> before and after studies, comparative trials, randomised trials

14. Randomized Control Design (RCT) ”Gold standard” in epidemiological research Steps: Two or more groups assigned by randomization: exp. vs control Makes study groups comparable: Controls for confounding (known and unknown), prevents selection bias Take baseline measure on all groups Give different treatments Measure outcome Quantitative Research 13

15. Scheme Group 1: O1 X1 O2 Group 2: O1 X2 O2 Where O= Observation Where X= Treatment Quantitative Research 14

16. Therapeutic Study population Patients with disease Objectives Cure patients Diminish symptoms Prevent recurrence of disease/risk of death Preventive Study Population Population at risk Objectives Reduce the risk of developing disease Types of intervention studies

17. Phasing intervention research • Phase I: healthy volunteers, testing for harm, etc • Phase II: small groups of patients, effective dose, interaction. • Phase III: randomized clinical trial (RCT), protocolled, sufficient sample size • Phase IV: systematic registration of harm

18. Design RCT I O P C

19. Study population Decide inclusion and exclusion criteria • Example of inclusion criterion: “All newly diagnosed head and neck cancer patients with small cell carcinoma that smoke” • Example of exclusion criterion: “Those who are terminally ill, mentally or cognitively unstable, pregnant, etc.” Exclusion criteria control for errors Quantitative Research 18

20. Randomization (1) Assigned to groups by method similar to “flipping a coin” If randomization works, groups will be the same/comparable The larger the sample, the greater the likelihood of equal groups Results should show that the demographic characteristics between groups are similar If groups are similar, do not need to control for baseline variables Quantitative Research 19

21. Randomization (2) Some times cannot randomize people (e.g., cross-contamination or “system” interventions) Can randomize hospitals, or units instead Once randomized, always randomized Subjects are treated as part of that group, even if they die, are lost to follow up, or withdraw Quantitative Research 20

22. Blinding Un-blinded: Everyone knows treatment Single Blinded: Researcher or patient does not know treatment Double Blinded: Neither researcher or patient knows treatment Why blinding? • Many people believe they feel better if they are given something: placebo effect • (Unintended) influence of researcher/person who takes measurements Quantitative Research 21

23. Intervention/Treatment Treatment versus placebo Treatment versus standard of care Treatments should be made to be as same as possible • For example, new drug versus sugar pill Quantitative Research 22

24. Intervention - fidelity checks Did the investigator make sure the intervention was delivered as it was supposed to be delivered? Becomes more of an issue when having multiple sites with multiple providers of the intervention Important to provide training and booster sessions and making spot checks Quantitative Research 23

25. Outcome variable The outcome variable should be the same as pretreatment measure Could have short-term and long term outcomes Intention to treat analysis Quantitative Research 24

26. RCT appraisel

28. Pros and cons of the RCT Advantages: • Prevention of bias: unbiased distribution of confounders • High level of evidence Disadvantages: • Expensive: time and money • Volunteer bias • Experimental circumstances ≠ circumstances in reality: external validity • Ethically problematic at times

29. Other designs Factorial designs Cross-over-trial

30. Four Arm Study Example: Factorial Design • High vs low intensitystrength training • Vitamin D supplements (1200IU/d) vs placebo in combinationwithstrength training 2 x 2 RCT n=220 knee OA + vitDdeficiency • High intensitystrength training + vit D • High intensitystrength training + placebo • Low intensitystrength training + vit D • Low intensitystrength training + placebo 1 3 2 4 Quantitative Research 29

31. Randomized Cross-Over Design A controlled trial where each participant receives both therapies, in randomized order Give each group the other treatment • O1 X1 O2 washout O3 X2 O4 • O1 X2 O2 washout O3 X1 O4 Quantitative Research 30

32. Randomized Cross-Over Design A controlled trial where each participant receives both therapies, in randomized order Give each group the other treatment • O1 X1 O2 washout O3 X2 O4 • O1 X2 O2 washout O3 X1 O4 Quantitative Research 31

33. Randomized Cross-Over Design Advantages: all participants receive treatment statistical tests assuming randomisation can be used blinding can be maintained Disadvantages: washout period lengthy or unknown cannot be used for treatments with permanent effects Quantitative Research 32

34. Summary

35. Different study designs

36. Reporting statements CONSORT for randomised controlled trials STARD for diagnostic accuracy studies STROBE for observational studies PRISMA for systematic reviews of trials MOOSE for meta-analyses of observational studies

37. Research question and design of your own Kneemo study?