2. CASE CONTROL STUDY

3. 8/21/20152 Case-control study Exposure Disease (+) ? -------------------------------------------- Exposure Disease (-) ? -------------------------------------------- Investigator

4. 8/21/2015B.Shakiba3 Case control study

5. 8/21/2015B.Shakiba4 Type of studies

6. Exposed Unexposed Source population

7. Cases Exposed Unexposed Source population

8. Cases Exposed Unexposed Source population Sample Controls

9. Cases Exposed Unexposed Source population Controls = Sample of the denominator Representative with regard to exposure Controls Sample

10. CasesControls Exposedab Not exposedcd Totala + cb + d % exposeda/(a+c)b/(b+d)

12. CASE-CONTROL STUDIES  Basic Idea: - Cases – Should represent all cases in the population - Controls – Should represent all persons without disease in the population

13. CASE-CONTROL STUDIES Population Sample

14. Case control study Disease Controls Exposure ???? Retrospective nature

15. Two types of case control studies,Exploratory New disease New risk factors Several exposures "Fishing expedition“,Analytical Precise a single hypothesis Dose response

16. REVIEW  A design used to assess the relationship between the exposure to a risk factor and the development of a disease  It compares the exposure distributions between the groups of patients with and without the disease.  It typically uses only a fraction of the subjects in the non-disease group.

17. Randomized Clinical Trials vs. Case-Control Studies Exposure No Exposure Randomiz ation Patients in Baseline State Patients with Disease Patients without Disease

18. Characteristics of the Design  Retrospective  No randomization  Population at risk is often undefined  Ascertainment of exposure history

19. Implementation a Case-Control Study: Practical Issues  Selecting a study base representative of the intended population  Defining the disease  Choosing the cases and controls  Exclusion criteria  Ascertainment of exposure

20. Selection of the Study Base  Hospital based case-control studies: The study base is the collection of clinical records of the participating hospitals. - Berkson’s Bias: Cases and controls experience different hospital admission rates.  Population based case control studies: The Study base is the collection of subjects who would become cases if they develop diseases. - Neyman’s Bias: Case group not representative of the intended population.

21. Diagnostic Criteria and Case Selection  Diagnostic criteria: unambiguous definition under equal diagnostic surveillance.  Sources of cases: 1.Persons with the disease seen at a care facility in a specified period of time. 2.Persons with the disease in a more general population in a period of time.

22. Selection of Controls Basic Principles  True Representation of the Study Base: The controls should be selected so that they truly represent the distribution of exposure in the study base from which the cases are selected.  Comparable Accuracy: There should be no differential misclassification between the two groups.

23. CASE-CONTROL STUDIES Sources of Cases and Controls  Population-Based - Cases – from Registry (fed by population) - Controls – from General Population  Hospital-Based - Cases - selected group that made it to hospital - Controls – as above

24. Selection of Controls: Sources  The controls should be drawn from the population of which the cases represent the affected individuals.  Sampling Frames: 1.Population of an administrative area (eg. HMOs) 2.Hospital patients 1.Difference with target population 2.Cost effective 3.Relatives of the cases (spouses and siblings) 4.Associates of the cases (neighbors, co- workers, etc)

25. Matching  Frequency matching  Individual matching

26. Matching  Advantages: - Sometimes the only way of control of some confounding in certain situations - Increasing power - Straightforward way to obtain a comparable group

27. Matching  Disadvantages: - Some time impossible - Association between matching variable and the outcome can’t be assessed - Not possible to assess theadditive interaction between matching variable and exposure - Increased int validity may result in reduced ext. validity - Considering OVERSTIMATION: not highcorrelation between the variable of interest and matching variable eg: matching ethnic background - No statistical power is gained if the matched variable is a weak confounder

28. Selection of Controls: Sampling Schemes  Total population – no sampling  Random and systematic sampling  Matching – deliberately select the controls in such a way as to make them similar to the cases with respect to certain confounding variables.  Multiple control groups.

29. Multiple controls  Similar  different

30. Exclusion Criteria  Exclusion criteria should not alter the exposure rate in one of the two groups.  Examples: 1.Low-level lead exposure and mental retardation-children with lead related diseases were excluded from the control group; 2.Reserpine and breast cancer-patients with thyrotoxicosis, renal disease, and cardiovascular diseases were excluded from the control group.

31. Information on Exposure  The most common sources of information on exposure are patients (or parents, in the case of children).  Other sources include relatives, hospital records, employment records, etc.  When information is obtained via interviews, recall bias is often a concern.

32. Information on Exposure: Comparability and Validity  Comparability: If the inaccuracy in exposure reporting affects the two groups to a different degree, the study may yields questionable conclusions.  Validity: The information on exposure reflects the true level of exposure.

34. Time - DZ E E E E Case-Control Study

37. Bias?

38.  Long latency period of disease  Controls should be selected independent of exposure  Controls: neighbor Other hospital wards Phone book Friends

39. Advantage and disadvantages

40. 8/21/201539 Case control studies  epidemiologists use them to study a huge variety of associations.  more frequently than other analytical studies

41. 8/21/201540 Case control studies Advantages:  Rare diseases  Several exposures  Long latency  Rapidity  Low cost  Small sample size  No ethical problem  Efficient, cost-effective for rare outcomes

42. Strengths of the Case- Control Study Design  It is less constrained by the natural frequency of the disease  It greatly shortens the waiting time required by cohort studies  It can be used when an RCT is not logistically or ethically feasible. Efficiency, low cost, fewer practical restrictions

44. 8/21/201543 Case control studies Disadvantages:  Selection bias  Measurement of exposure information  Control of confounding factors  Not suitable for rare exposure  ? Sequence of events ?  Only one outcome  Does not yield incidence or relative risk (although in some cases these can be inferred using external information)  BIAS

45. Limitations of case-control studies  Cannot compute directly relative risk  Not suitable for rare exposure  Temporal relationship exposure-disease difficult to establish  Biases +++ - control selection - recall biases when collecting data  Loss of precision due to sampling

46. Weaknesses of the Case- Control Study Design  Less well defined target population  Concerns of various types of selection biases  Lack of causal interpretation due to the existence of confounding factors  No absolute measure for the exposure effect on the disease

47. Effects

48. INCIDENCE or PREVALENCE DISEASE or EXPOSURE Intuitively if the frequency of exposure is higher among cases than controls then the incidence rate will probably be higher among exposed than non exposed.

49. CasesControls Exposedab Not exposedcd Totala + cb + d % exposeda/(a+c)b/(b+d) Distribution of cases and controls according to exposure in a case control study

50. PhysicalMyocardial activityInfarctionControls >= 2500 Kcal190230 < 2500 Kcal176136 Total366366 % exposed51.9%62.8 % Distribution of myocardial infarction cases and controls by amount of physical activity

51. WaterCases Controls Consumption YES 150 ? NO 50 ? Total 200 200 Volvo factory, Sweden, 3000 employees, Cohort study 200 cases of gastroenteritis

52. Probability that an event will happen Odds= Probability that the even will not happen Probability that an event will happen Odds= 1 - (Probability that the event will happen)

53. Case control study CasesControls Exposedab Not exposedcd Totala + cb + d Odds of exposure among cases = Probability to be exposed among cases Probability to be unexposed among cases a / (a+c) Odds Ecases =------------ = a / c c / (a+c) Odds of exposure among controls = Probability to be exposed among controls Probability to be unexposed among controls b/ (b+d) Odds Econtrols = ------------ = b / d d/ (b+d) a/c OR = ---- = ad / bc b/d

54. CASE-CONTROL STUDIES  BASIC IDEA  Is the risk factor more common among than ?  Is the risk factor more common among cases than controls? (+) (-) CaseControl RF (+) 50 20 RF (-) 50 80 100 RF PREVALENCE FOR CASES 50/100=50% RF PREVALENCE FOR CONTROLS 20/100=20% PREVALENCE RATIO = = 2.5 PREVALENCE RATIO = 50% / 20% = 2.5

55. CASE-CONTROL STUDIES  BASIC IDEA  Is the risk factor more common among than ?  Is the risk factor more common among cases than controls? ODDS FOR CASES 50:50 = 1 ODDS FOR CONTROLS 20:80 = 0.25 ODDS RATIO = 50:50/20:80 = 1/0.25 = 4 (+) (-) CaseControl RF (+) 50 20 RF (-) 50 80

56.  RR isn’t possible to calculate in case control study  OR is calculated  OR is representative of RR if: - Cases are representative - Controls are representative - Disease prevalence is rare

57. CASE-CONTROL STUDIES a b c d (+) (-) Exp Disease (+) (-) For rare diseases, a is small compared to b, and c is small compared to d. So..... a/(a+b) c/(c+d) Relative Risk = a/(a+b) c/(c+d) a/b c/d ≈ a/c b/d = = Odds Ratio

58. CASE-CONTROL STUDIES  Method: Population-based  Prospective case-control   Cases: All incident cases of childhood (<15 yo) cancer in Denver registry, 1976-1983  Controls: Random-digit dialing match on sex, age ± 3y

59. CASE-CONTROL STUDIES 11 31 34 175 Results Electric Blanket (+) (-) Case Control Brain Cancer Case Odds: 11/34 = 0.32 Control Odds: 31/175 = 0.18 Odds Ratio: 0.32/0.18 = 1.8 (95% CI = 0.9 – 4.0)

60. Analytical Issues  Association vs Causal relationship.  Adjustment of confounders: 1.Matching 2.Model based adjustment (regression, etc) 3.Propensity score method 4.A common limitation of the adjustment: cannot account for the effects of the unobserved confounders.

61. Assessment of the Effect of the Exposure CaseNon-caseSubtotal ExposedABA+B Not exposedCDC+D SubtotalA+CB+DA+B+C+D

62. Relative Risk and Odds Ratio  For a RCT, we report the relative risk:  For a Case-Control study, we report the odds ratio:  RR and OR:

63. Final Thoughts  Thoughtful design and careful implementation.  Reducing biases of various kinds.  The workhorse of the case-control data analysis is logistic regression.  Reporting a case-control study.

64. Nested case control