Study design used in pharmacoepidemiology

This is the topic of study design used in pharmacoepidemiology for pharmacy students.

1. Study design used in Pharmacoepidemiology สอน โดย อ. กมลวรรณ ตัน ติพิวัฒน สกุล ได้รับการสนับสนุนจาก อ. มนทรัตม์ ถาวรเจริญทรัพย์ 1

2. Study design used in Pharmacoepidemiology Example: Vitamin C and Common cold Observational study Case report/ case series Ecological study Cross-sectional study Case control study Cohort study Experimental study Randomized clinical trial (RCT) Parallel design Cross-over design Factorial design Cluster randomized trial Outline 2

3. What is “ Pharmacoepidemiology ”? Pharmacoepidemiology : The application of epidemiological principles and methods to the study of drug effects in human population. The study of the use of and the effects of drugs in large number of people . A new science that uses principles of epidemiology in quantifying adverse drug events , pattern of drug use , and drug efficacy in a large population . Pharm World Sci , 1995; 17(3);61-65. Strom BL, Pharmacoepidemiolgy 3 rd Edition, 2001 J clin pharmacol 2006; 46; 6-9 What is being used? How it being used? (pattern of use including how much, where, when and by whom Why it being used (Reasons for drug-taking behavior/ compliance)

4. Study design Yes No Randomization? Yes No Comparison group? Yes No Direction? E O E O E & O at the same time Individual assignment Group assignment 4

5. Level of evidence High Low 5

6. Non-experimental study Experimental study Few ethical constraint Ethical constraint Easier to recruit and enroll subjects Harder to recruit, enroll and follow subjects Can be relatively quick May take long time Less expensive Expensive Lack of control of confounding / Prone to bias Clearer interpretation of causal relationship Comparing Non-experimental VS Experimental design 6

7. Case report: Case report is simply report of single patient. A case report describes a single patients who was exposed to a drug and experiences a particular, usually adverse outcome. Case series: Case series are collections of patients, all of whom have a single exposure, whose clinical outcome are then evaluated and described. Alternatively, case series can be collections of patients with a single outcome, looking at their antecedent exposures. Observe 100 women aged less than 50 years old, who suffer from a pulmonary embolism, and note that 30 of them had been taking oral contraceptive Case report/ case series 7

8. Advantage Disadvantage Cheap and easy method for generating hypothesis No control group No control over confounding Cannot be used for proven-causal-effect relationship Case report/ case series 8

9. Ecological study Ecological study (Analyses of secular trends) examine trends in an exposure that is a presumed cause and trends in a disease that is presumed effect and test whether the trends coincide. This trends can be examined over time or across geographic boundaries . analyze data from a single region and examine how it changes over time, analyze data from a single time period and compare how the data differ from region to region .

10. Ecological study Situation: The unit of analysis is a group, the number of exposed persons and the number of cases is known for each group, but the number of exposed cases is not known . Warning! The ecologic fallacy. Results from making a causal inference about an individual phenomenon or process of the observation of groups.

11. Cross-sectional study 11 Gather data on Exposure and Disease Cross-sectional study examines the relationship between disease/outcome and other variables of interest as their exists in a defined population at 1 particular time. In Cross-sectional study, study population are commonly selected without regard to exposure or disease status

12. Outcome status Yes No Exposure status Exposed A B A+B Non-exposed C D C+D A+C B+D A+B+C+D = N 2 X 2 Table: 12 Case-control study cohort study Cross-sectional study

13. Outcome status Yes No Exposure status Exposed A B A+B Non-exposed C D C+D A+C B+D A+B+C+D = N 2 X 2 Table: Cross sectional study 13 Cross-sectional study Calculation: Prevalence of outcome in person with exposure: A/(A+B) Prevalence of outcome in person without exposure: C/(C+D)

14. Muscular complaint Yes No Exposure status Lipid -lowering Drug (LLD) 106 133 239 No lipid-lowering Drug 305 487 792 411 1031 2 X 2 Table: Cross sectional study 14 Cross-sectional study Calculation: Prevalence of muscular complaint in person using LLD: 106/239 Prevalence of muscular complaint in person not using LLD: 305/792

15. Advantage Disadvantage Permit determination of various characteristic of a population Not useful for rare disease or disease with short duration Generate hypothesis for exposure-disease relationship Not infer temporal sequence between exposure and disease Example: Cigarette and infertility Diazepam and stress Cheap and easy to conduct Cause-effect relationship is tenuous Cross-sectional study 15

16. Case control study 16 Present Past Retrospective Study that compare cases with the disease to control without the disease, looking for differences in antecedent exposure. (Retrospective)

17. Sources of cases: Ideally- all incident cases in a defined population in a specified time period. Source of case identified: National health survey, specific reporting systems (cancer registry, birth defect registry ) Specific criteria based on a combination of sign and symptoms MI: Criteria A = Chest pain “Silent heart attack” was excluded People with other condition that produce chest pain was mistakenly included. Prevalent VS incident case: study cause of disease one prefer incident case because they usually interested in factors that lead to developing of disease rather than factors that affecting the duration of disease. Selection of case

18. Source of controls: Ideally- should have the same characteristics as the cases except for the exposure of interest. Source of controls: Population control, hospital control, friends, family, spouse. Etc. Selection of control

19. Source of exposure: Interview, self-administered questionnaire, pharmacy data, etc. Information about exposure must be accurate Information on dose and duration is important! The definition of exposure must be appropriate for causal inference “Ever used” as exposed might not be accurate Exposure identification 19

20. Outcome status Yes No Exposure status Exposed A B A+B Non-exposed C D C+D A+C B+D A+B+C+D Example: 2 X 2 Table (case-control) 20 Case-control study Measure of association Odd ratio (OR) = AD/BC

21. Advantage Disadvantage Quick and easy Exposure-disease time relationship is not clear Inexpensive Does not provide direct estimate of risk Particularly useful for rare disease Possibility of introduction of bias, relies on recall, incomplete control of extraneous factor Causal-relationship is not established Case-control 21

22. Cohort study 22 Present/future Past/ Present Prospective

23. Past Present Future Prospective Cohort Retrospective Cohort Cohort: Direction and timing 23 Study begins Study begins Both prospective cohort and retrospective cohort is prospective study!

24. Outcome status Yes No Exposure status Exposed A B A+B Non-exposed C D C+D A+C B+D A+B+C+D 2 X 2 Table: Cohort study 24 cohort study Relative Risk (RR) = A/ (A+B) C /(C+D) Risk difference (RD) = A/(A+B) – C/(C+D) NNT = 1/RD Measure of association

25. Advantage Disadvantage Direct determination of risk Take long time Strong evidence of exposure-disease association Difficult to implement and carry out Can study rare exposure Expensive Chance of loss to follow-up Change in exposure status may occur Need large sample to study rare disease High chance of c onfounding especially confounding by indication Cohort study 25

26. RCT study 26 future Present Prospective R* * Randomization

27. Cohort VS RCT 27 Not randomly allocated (e.g. self selected) Randomly allocated/ Randomization

28. Type Characteristic Parallel VS Crossover Parallel: Each group receive one treatment. Treatments are administered concurrently Crossover: Each group receive all treatment one after another. Treatments order differs for each group. Washout period may intervene between treatment. Simple VS factorial Simple: Each group gets one treatment Factorial: Each group gets two or more treatments Parallel VS Crossover VS Factorial 28

29. Parallel VS crossover design 29 Washout Washout Phase I Phase II Parallel design Crossover design

30. HIV-Infected Yes No Exposure status Zidovudine 9 141 150 Placebo 31 118 149 Example:2 X 2 Table (RCT) 30 RCT study RR Zidovudine VS Placebo = (9/150) / (31/149) Adapted from N Eng J Med 1994:331:1173-1180

31. Example: crossover design 31 Washout (2 weeks) Washout Phase I (3 weeks) Phase II ( 3 weeks)

32. Parallel Crossover Each group receive only one treatment. (Randomization of treatment group) Each group receive both treatment but in different order (Randomization of sequence) Many situations can only be studied in parallel such as fatal outcome Cannot studied some outcomes such as fatal outcome Typically need large N Can have fewer subjects and be convincing Dropout reduce power but not drastically Dropout reduces power dramatically No carryover effect Carryover effect may occur if washout period is not enough Can take long time Not appropriate to study long term Parallel VS Crossover 32

33. Factorial design: Example 33 Objective: to analyse effect - Tocopheral and - carotene of and on incidence of cancer and mortality R GR 1 GR 2 GR 3 GR 4

34. Outcome status Yes No Exposure status Exposed A B A+B Non-exposed C D C+D A+C B+D A+B+C+D 2 X 2 Table: RCT 34 R C T study Measure of association Relative Risk (RR) = A/ (A+B) C /(C+D) Risk difference (RD) = A/(A+B) – C/(C+D) NNT = 1/RD

35. Cluster randomized trials are experiments in which social units or clusters rather than individuals are randomly allocated to intervention groups 35 Cluster randomization study

36. Example: Cluster randomization study

37. Ethics in experimental study Uncertainty in experimental study Can’t give a harmful therapy to treatment group. There should be a reasonable chance that the benefits outweigh potential risks. Can’t withhold a beneficial treatment from control group. (Use a “standard care” control group as an alternative) Informed consent Monitoring boards

38. Informed consent Participants should understand: Type of study What is requested? Risks and benefits The concept of randomization and blinding The right to withdraw The right to be informed of relevant findings

39. Randomization Concept : All individual have same and independent chance of being allocated to any of the treatment groups; produces comparable groups . However, a llocation ratios may vary (1:1, 1:2, 1:3) . Goal: To prevent bias related to selection. Balances treatment group on known and unknown characteristics (minimized confounding)

40. Randomization Desirable properties of randomization method Ease of implementation Small imbalance in sample size Difficult to guess next assignment The following is not examples of randomization: Assigning every other patient to treatment A and others to B . Assigning equal numbers to treatments A and B based on what is best for patient. Assigning A to patients recruited from one source and B to patients from another source.

41. Basic concept: Coin tossing (e.g. Head = A, Tail = B) but clumsy and time consuming ! Using table of random digit (0-9). Each digit occurs on average the same number of times, there is no discernible pattern of digit values and the table present digits in pairs merely to help the user in scanning across the page. Simple randomization

42. For 2 treatments (1:1) assign A for digits 0-4 B for digits 5-9 Hence the numbers in the top row of table 0 5 2 7 8 4 3 7 4 1 6 8 3 8 5 1 5 6 9 6 produce a list: A B A B B A A B A A B B A B B A B B B B For 3 treatments ( 1:1 :1 ) assign A for digits 1-3 B for digits 4-6 C for digits 7-9 and ignore 0 produce a list: For 2 treatments (2:1) : How? Simple randomization

43. The advantage : Simple, each treatment assignment is completely unpredictable, the probability theory guarantees that in the long run the numbers of patients on each treatment will not be radically different. The disadvantage: High chance of imbalance in small trial. Simple randomization

44. Random Permuted Blocks Suppose we have T treatments, then for each block of kT patients we produce a different random ordering of k assignments to each treatment. For 2 treatments, blocks of 2 patients assign AB for digits 0-4 BA for digit 5-9 Then, the numbers 0 5 2 7 8 AB BA AB BA BA Block randomization

45. For 3 treatments, blocks of three patients ABC for Digit 1 ACB for Digit 2 BAC for Digit 3 BCA for Digit 4 CAB for Digit 5 CBA for Digit 6 Ignore 0 and 7-9 So, the number 0 5 2 7 8 4 Produce list - CAB ACB - - BCA Block randomization

46. Disadvantage : At the end of each block, a clinician who keeps track of previous assignments could predict what the next treatment would be. If the block size is large , the serious mid-block inequality might occur. Example : Block size of 4 if the first three patient received A, B, and B then the 4 th patient will receive A . Block randomization

47. Unblinded = Open trial Blinding: the treatment assignment is not known to certain persons. Single-blinded study, the treatment assignment is unknown to the patients Double-blinded study, the treatment assignment is unknown either to the patients or to their physicians Triple-blinded study , the treatment assignment is unknown to the patient, the physician, and the committee monitoring response Blinding

48. Goal: Prevention of bias (often in outcome assessment) Issues: Ethics, breaking the blind Difficulty blinding some treatment Drug with side effects: estrogen and bleeding Lifestyle changes, surgery procedure Assess the blinding whether the blinding still remains. Blinding

49. Matched placebo: id entical in all respects to the active oral drug except that the active ingredients is absent. Particular features requiring matching are the colour, taste, texture, shape, size. Double dummy: When 2 or more active drugs are being compared.Double dummy may be the option: Each active ingredient has a placebo identical to it. Each patient would take one active and one placebo pills. (Not reasonable when several active drugs are being compared.) Blinding

50. Drug A ( b. i .d .) vs Placebo: How? Different dose schedule: How? A = new drug (sustained release) 100 mg o.d, B = conventional drug 100 mg t. i .d. Put in identical capsule: new drug, conventional, and placebo: then take tid in both group. Or using double dummy; Breakfast Lunch Evening Gr A Real A B dummy B dummy B dummy Gr B A dummy Real B Real B Real B Blinding

51. Systematic review Systematic review: is a summary of the medical literature that Use explicit methods Is based on a through literature search Performs a critical appraisal of individual studies Synthesize the world literature on a specific issue Use statistical techniques to combine data from valid studies (meta-analyses) Systematic review may or may not include meta-analysis Sackett DL, Strauss, S.E., Richardson, W. et.al. Evidence based medicine: How to practice and teach evidence based medicine” London” Churchill-Livingstone.2002.

52. Why systematic review? Expanding volume of published literature Different or controversial results from studies of the same topic

53. Systematic review VS Narrative review Narrative Systematic Informal and subjective methods to collect and interpret studies Formal and objective method to collect and synthesis the result from studies Not always conduct in extensive search Extensive search Rarely explicit about how they select the study. Tend to be selective in citing reports that reinforce their preconceived ideas Using explicit method with clear and reproducible eligibility criteria to select the study for review Less rigorous critical appraisal Rigorous critical appraisal High risk of bias Minimal bias

54. Meta-analysis Meta-analysis: “.. A quantitative approach for systematically combining the results of previous research in order to arrive at conclusion about the body of research. Petitti D.B., Meta analysis, decision analysis, and cost-effectiveness analysis. New York: Oxford University Press, 1994.

55. Thank you for your attention 55