1. BIAS

2. ExplanationAssessment Strategy Random variabilityEstimation of precision (95% confidence interval), and testing (p) ConfoundingExperimental design; adjustment BiasQuality assurance and Quality control Causal relationshipEliminate alternative explanations; Hill’s criteria Possible Explanations for an Association Between a Risk Factor and a Disease

3. 1. Definition of Bias

4. TRUTH Study Results Frequency Average of Results DISTRIBUTION OF AN INFINITE NUMBER OF STUDIES Average bias Statistical definition of bias: When the average value of the association measure obtained from an infinite number of studies is not the true value

5. TRUTH Study Results Frequency Average of Results DISTRIBUTION OF AN INFINITE NUMBER OF STUDIES Difference between validity and representativeness: 1. Biased study Biased, but “representative”

6. Frequency Study Results DISTRIBUTION OF AN INFINITE NUMBER OF STUDIES TRUTH Average of Results Valid, but not “representative” Difference between validity and representativeness: 2. Valid (unbiased) study

7. Epidemiological Definition of Bias Last J: A Dictionary of Epidemiology, ed. by J. Last, 3rd Edition, IEA “Deviation of results or inferences from the truth, or processes leading to such deviation. Any trend in the collection, analysis, interpretation, publication, or review of data that can lead to conclusions that are systematically different from the truth.”

8. 2. Main types of bias a. Selection bias b. Information bias

9. Diseased Exposed Healthy Exposed Diseased Unexposed Healthy Unexposed Reference Population Study Sample Selection Bias: One group (cell) in the population (e.g., exposed cases) has a > likelihood of inclusion in study.

10. “Gold Standard”: Total Population Study A. Hypothetical Case-Control Study Including All Cases and All Non-Cases of a Reference Population 50:50 = 1:1180:720 = 1:4 (50/50) ÷ (180/720) = 4.0 Unbiased exposure odds in cases and controls Unbiased relative odds

11. 100 x 0.50= 50 25 18 72 900 x 0.10= 90 25:25 = 1:118:72 = 1:4 (25/25) ÷ (18/72) = 4.0 Unbiased exposure odds in cases and controls Unbiased relative odds B. Hypothetical Unbiased Case-Control Study Including 50% Cases and 10% Non-Cases of a Reference Population Unbiased Study Based on Samples

12. Example of Selection Bias Without Compensation Bias 50 x 0.60= 30 50 x 0.40= 20 180 x 0.10= 18 720 x 0.10= 72 100 x 0.50= 50 900 x 0.10= 90 30:20 = 1.5:1.018:72 = 1:4 (30/20) ÷ (18:72) = 6.0 Biased exposure odds in cases Unbiased exposure odds in controls Biased relative odds Conclusions C. Hypothetical Biased Case-Control Study Including 50% Cases and 10% Non-Cases of a Reference Population

13. Selection Bias: Compensating Bias D. Hypothetical Case-Control Study Including 50% Cases and 10% Non-Cases of a Reference Population with Compensation Bias 100 x 0.50= 50 900 x 0.10= 90 50 x 0.60= 30 50 x 0.40= 20 180 x 0.13= 24 720 x 0.09= 66 30:20 = 1.5:1.0 24:66 = 1.0:2.7 (30/20) ÷ (24/66) = 4.1 Exposure odds equally biased in cases and controls rounding Comparability of selection processes Unbiased Relative Odds

14. True Odds: In cases: 1:1 In controls: 1:4 Bias: In cases: [1.5:1]  [1:1]= 1.5 In controls: [1:2.7]  [1:4]  1.5 = True OR

15. Example of Compensating Bias Case-control study of breast cancer identified in a population-based screening program Controls? - Random sample of total population? = screened women Women who have normal mammograms, or all women undergoing screening Cases Controls Reference Population* *Study Base

16. Example 1* Hypothesis: HBV is associated with aplastic anemia (AA) Cases: AA patients hospitalized at the Johns Hopkins Hospital Hematology Division (a referral Division) Controls: Patients with non-malignant diseases admitted to the Johns Hopkins Hospital PROBLEM?? *Based on a study by Szklo et al TWO EXAMPLES OF UNCOMPENSATED BIAS

17. Patients (cases) hospitalized with AA at the Johns Hopkins Hospital in Baltimore, USA, usually undergo bone marrow transplant To be eligible for bone marrow transplant, patient is more likely to: –Come from large family (where a genetically-matched donor is more likely to be found), and –Have medical insurance (as bone transplants are very expensive), and thus belong to a higher S.E.S. than control patients. THUS, VARIABLES RELATED TO FAMILY SIZE AND TO SOCIO-ECONOMIC STATUS  SELECTION FACTORS THE CHOICE OF CONTROLS DID NOT FOLLOW THE BASIC PRINCIPLE: THAT CASES AND CONTROLS SHOULD BE CHOSEN FROM THE SAME STUDY BASE!

18. Coffee Drinking and Cancer of the Pancreas (MacMahon et al, New Eng J Med 1981;304:630) Cases: Patients newly diagnosed with pancreatic cancer admitted to 11 Boston and Rhode Island hospitals during 1974-1980 (n= 369) Controls: Patients under the care of the same MD’s in the same hospitals, interviewed at the same time as the cases (i.e., cases and controls were matched by attending physician, hospital, and timing of interview). Controls mostly had gastrointestinal conditions, or cancers other than pancreatic and biliary tract (n= 644) Example 2 TWO EXAMPLES OF UNCOMPENSATED BIAS

19. Odds ratios for cancer of the pancreas according to coffee drinking and smoking Smoking Coffee intake (cups/day) 01-23+ Never1.02.13.1 Ex-smokers1.03.12.3 Current smokers 1.01.83.8 Total1.01.82.7 (Adapted from MacMahon et al, New Eng J Med 1981;304:630) Problem?

20. Diseased Exposed Controls Diseased Unexposed controls Reference Population Study Sample

21. 2. Types of Bias a.Selection bias b. Information bias and misclassification

22. Diseased Exposed Healthy Exposed Diseased Unexposed Healthy Unexposed Reference Population Study Sample Exp Unexp Information bias: misclassification of exposure information in cases and controls Cases Controls a b c d

23. Examples of Information Bias Exposure Identification Bias –Recall Bias –Interviewer/Observer Bias Outcome Identification Bias – Observer Bias – Respondent Bias

24. Result of Information Bias: Misclassification Differential Non-differential

25. Example of Recall Bias in a Study of Melanoma (Weinstock et al, Am J Epidemiol 1991;133:240-5) Healthy cohort: Data collection (questionnaire) on tanning ability Time (cohort follow-up) Case Loss

26. 0.7Odds Ratio 15525Medium, average, deep or dark tan (“unexposed”) 799No tan or light tan (“exposed”) CoCaCoCaTanning Ability Post- melanoma diagnosis Pre- melanoma diagnosis +6 +2 15 19 77 157 1.6 Example of Recall Bias in a Case-Control Study (Ca-Co) (Weinstock et al, Am J Epidemiol 1991;133:240-5)

27. To investigate the validity of self-reported acquired immunodeficiency syndrome (AIDS) among women enrolled in a prospective study of human immunodeficiency virus (HIV) infection, the authors compared the self-reported occurrence of AIDS-specific diagnoses with AIDS diagnoses documented by county AIDS surveillance registries. (Hessel et al, Am J Epidemiol 2001;153:1128-33) Example of nondifferential misclassification “test” “gold standard”

28. Sensitivity and Specificity of Self-Reported Diagnosis of Esophageal Candidiasis in AIDS patients Sensitivity: 46% Specificity: 84% (Adapted from: Hessol et al, Am J Epidemiol 2001;153:1128)

29. Relationship between AIDS and esophageal candidiasis: “Gold Standard” 1,000500Total 1.0995480Absent 8.3520Present Odds Ratio Normal controls AIDS cases Esophageal candidiasis

30. Definitions of Sensitivity and Specificity Used for the Evaluation of Misclassification Sensitivity –Proportion of all truly infected (exposed) individuals correctly classified by the study Specificity –Proportion of all truly uninfected (unexposed) individuals correctly classified by the study

31. Relationship between AIDS and esophageal candidiasis ascertained by questionnaire with sensitivity= 46% and specificity= 84%. Assume non-differential misclassification* 1,000500Total 1.0995480Absent 8.3520Present Odds Ratio Normal controls AIDS cases True Results Truth in Cases 50048020Total Absent Present In StudyAbsentPresent Self-report in Cases *Same sensitivity and specificity values for cases and controls

32. Relationship Between AIDS and esophageal candidiasis ascertained by questionnaire with sensitivity= 46% and specificity= 84% 1,000500Total 1.0995480Absent 8.3520Present Odds Ratio Normal controls AIDS cases Self-report of esophageal candidiasis True Results: Truth in Cases 50048020Total Absent Present In StudyAbsentPresentSelf-report in cases 11 9 403 7786 414

33. Relationship Between AIDS and esophageal candidiasis ascertained by questionnaire with sensitivity= 46% and specificity= 84% 1,000500Total 1.0995480Absent 8.3520Present Odds Ratio Normal controls AIDS cases True Results Truth in Cntrls 10009955Total Absent Present In StudyAbsentPresent Self-report in Controls

34. Relationship Between AIDS and esophageal candidiasis ascertained by questionnaire with sensitivity= 46% and specificity= 84% 1,000500Total 1.0995480Absent 8.3520Present Odds Ratio Normal controls AIDS cases Self-report of esophageal candidiasis True Results: Truth in Cntrls 10009955Total Absent Present In StudyAbsentPresentSelf-report in controls 3 2 836 159 161 839

35. Truth in Cases 50048020Total 41440311Absent 86779Present In StudyAbsentPresent Self-report in cases Truth in Controls 10009955Total 8398363Absent 1611592Present In StudyAbsentPresent Self-report in controls 1,000500Total 839414Absent 16186Present Normal controls AIDS cases Esophageal candidiasis Study Results: OR= (86/414) ÷ (161/839) = = 1.1 (True OR= 8.3!!) Rule: When there are two exposure categories, non- differential misclassification biases odds ratio or relative risk toward 1.0

36. Odds Ratios for Inaccurate Self-Reporting of AIDS (Hessol et al, Am J Epidemiol 2001;153:1128) *Simultaneously adjusted for all other variables using multiple logistic regression.

37. Examples of Differential Misclassification in a Case- Control Study. True Odds Ratio= 3.86; Prevalence of Exposure in Controls= 0.10 Exposure Ascertainment Odds Ratio SensitivitySpecificity CasesControlsCasesControls 0.900.601.00 5.79 0.600.901.00 2.22 1.00 0.900.701.00 0.700.904.43

38. Net result of misclassification: Regression Dilution Bias Example: Blood pressure as the exposure

39. Time SBP (mmHg) 120 140 Regression towards the mean: Random variability Measurement error Physiologic variability average

40. Time SBP (mmHg) 120 140 HypertensiveNormotensive Hypertensive Classification:

41. Time SBP (mmHg) 120 140 “Hypertensive” Normotensive Person A Person B Classified as:

42. 2. Types of Bias a. Selection bias b. Information bias and misclassification c. Mixed biases - Prevalence-Incidence bias - Temporal bias

43. Cross-sectional study If the prevalence is low: If exposure to the risk factor does not affect the duration of the disease after it starts: Risk Ratio

44. EXAMPLE: MYOCARDIAL INFARCTION Hypothetical numerical example Yearly incidence in persons older than 60: Men: 5% Women: 2% Survival after the acute event: Men: 20 years Women: 10 years Prevalence= Incidence × duration RR= 2.5

45. 2. Types of Bias a. Selection bias b. Information bias and misclassification c. Mixed biases - Prevalence-Incidence bias - Temporal bias

46. Type of estrogen No. of cases No. of controls Odds Ratios 95% confidence intervals None2743901.0Reference Conjugated56184.32.5, 7.5 Total339489 Number of Cases and Controls, and Odds Ratios for Endometrial Cancer According to Type of Estrogen Replacement Therapy, with 95% Confidence Intervals (Antunes et al, NEJM 1979)

47. Estrogen Use  Endometrial Cancer? OR Undiagnosed Endometrial Cancer  Bleeding  Estrogen Use  Diagnosed Endometrial Cancer ? Feinstein & Horowitz’ criticism Solution? Analyze only women who take estrogen prophylactically

48. Example of Temporal Bias: Relationship of Bypass Surgery to Physical Activity Cases who had bypass surgery Vs. controls without coronary heart disease Question: Do you exercise often now? “Yes”: Cases > Controls Question: Did you exercise often before your bypass surgery/before (date)? “Yes”: Cases < Controls  “Reverse Causality”

49. IS CONFOUNDING A BIAS? Sedentary life  Oral cancer alcohol Confounded relationship A DETOUR…

50. IS CONFOUNDING A BIAS? Sedentary life  Oral cancer alcohol unknown High risk marker  useful for secondary prevention Types of Association (Lilienfeld) Causal Statistical non-causal (“indirect”, due to confounding) Spurious or artifactual (due to bias) True

51. Is confounding a bias? Public health implications GoalType of evidence needed Primary prevention Prevention or cessation of exposure E.g.: saturated fat intake and atherosclerosis Causal association must be present, otherwise, intervention on risk factor will not affect disease outcome E.g.: if excessive fat did not cause atherosclerosis, a lower fat intake would not affect atherosclerosis risk Secondary prevention (screening) Early diagnosis via selective screening of “high risk” subjects E.g.: screening for hypertension in African- Americans Associations may be either causal or statistical (the latter must not be biased). In other words, the association may be confounded, but it is still useful for secondary prevention. E.g.: even if “race” is not causally related to hypertension (but confounded by SES, etc.), it could be a useful marker to detect individuals at higher risk for hypertension

52. 2. Types of Bias a. Selection bias b. Information bias and misclassification c. Mixed biases - Prevalence-Incidence bias - Temporal bias d. Biases in the evaluation of screening

53. Biases in the Evaluation of Screening Programs Lead Time Bias Selection Biases Referral Bias Length-biased sampling

54. Biases in the Evaluation of Screening Programs Lead Time Bias 

55. Biologic onset A Earliest point when diagnosis is possible B C Point when early diagnosis is made D Usual diagnosis based on symptoms F Outcome (e.g., death, recurrence) Detectable Preclinical Phase Survival after early diagnosis Natural History of a Disease (Adapted from Gordis, Epidemiology, 1996) Survival after usual diagnosis Lead Time

56. Survival 6 years 2000 1985 B 2000 1985 A Diagnosis Based on Symptoms 1994 1992 Early Diagnosis (Screening) Survival 8 years Lead time (2 years) LEAD TIME BIAS No protective effect: Survival B = Survival A + lead time = = 8 years = 6 years + 2 years

57. 2002 Survival:10 years 2000 1992 2000 1994 A B Survival: 6 years Protective effect: Surv (B) > Surv (A) + Lead Time = = 10 years > 6 years + 2 years Gain= 2 yrs Survival: 8 years Lead time Diagnosis based on symptoms Early Diagn.

58. 100% 70% Cumulative Survival Natural History of a Disease: Lead Time Bias (Adapted from Frank, Am J Prev 1985;1:3-9) Years after diagnosis 40% 5 years after usual diagnosis -2 5 years after early diagnosis= 3 years after usual diagnosis 3 5 1012 Lead Time bias

59. Lead time: Prevention and Correction Prevention: Use mortality instead of case-fatality Correction: Estimate lead time, and adjust for it –Examples Breast Cancer: 1 year Invasive Cervical Cancer: At least 10 years? Lung Cancer: Less than 1 year?

60. Biologic onset A Earliest point when diagnosis is possible BD Usual diagnosis based on symptoms F Outcome (e.g., death, recurrence) Detectable Preclinical Phase Estimation of Lead Time Step 1= Estimation of Duration of DPCP

61. Estimation of Lead Time Step 1= Estimation of Duration of DPCP Prevalence of the DPCP? Time 1 st exam 2 nd exam Incidence of the DPCP? Also, incidence of clinical cases

62. Biologic onset A Earliest point when diagnosis is possible BD Usual diagnosis based on symptoms F Outcome (e.g., death, recurrence) Detectable Preclinical Phase Estimation of Lead Time Step 1= Estimation of Duration of DPCP

63. Estimation of Lead Time Step 2= Estimation of Lead Time Time 1 st exam a. Prevalent Cases

64. Earliest point when diagnosis is possible BD Usual diagnosis based on symptoms Detectable Preclinical Phase Possible points in time when early diagnosis is possible

65. When screening exams are frequent, the lead time approximates the DPCP b. Incident Cases

66. Estimation of Lead Time Step 2= Estimation of the Lead Time Time 1 st exam Patient A Patient B 2 nd exam LEAD TIME DPCP b. Incident Cases

67. Biases in the Evaluation of Screening Programs Lead Time Bias Selection Biases 

68. Biases in Evaluation of Screening Selection Bias –Referral Bias (Volunteer Bias)  DEFINITION and SOLUTION? – Randomized Clinical Trial

69. Biases in Evaluation of Screening Selection Bias –Referral Bias (Volunteer Bias) –Length-Biased Sampling  – DEFINITION and SOLUTION?

70. Length-Biased Sampling (each horizontal line represents the DPCP for a case) Screeening Exam No. 1 Screening Exam No. 2 1 year Interval Cases

71. Lead-Time- Adjusted Five-Year Case-Fatality Rates Among Breast Cancer Patients (Shapiro et al, JNCI 1982;69:349-55) SCREENED STUDY GROUP

72. Lead-Time- Adjusted Five-Year Case-Fatality Rates Among Breast Cancer Patients (Shapiro et al, JNCI 1982;69:349-55) SCREENED STUDY GROUP Only valid comparison

73. Biases in Evaluation of Screening Selection Bias –Referral Bias (Volunteer Bias) –Length-Biased Sampling  – DEFINITION and SOLUTION? »Randomized Clinical Trials »Compare all individuals randomized to the “experimental” group with all individuals randomized to the control group