Ce diaporama a bien été signalé.
Nous utilisons votre profil LinkedIn et vos données d’activité pour vous proposer des publicités personnalisées et pertinentes. Vous pouvez changer vos préférences de publicités à tout moment.
CASE-CONTROL STUDY
Presentation by: Dr. Nidhi Singh
Moderator: Dr. Geeta Pardeshi
Department of Community Medicine
VMMC an...
CONTENTS
 Introduction
 Study Design
 Basic steps
 Bias in Case Control Study
 Strengths and Weaknesses of the study
...
CLASSIFICATION
EPIDEMIOLOGIC
METHODS
OBSERVATIONAL
STUDIES
DESCRIPTIVE
STUDIES
ANALYTICAL
STUDIES
ECOLOGICAL
CROSS-
SECTIO...
INTRODUCTION
 In contrast to DESCRIPTIVE STUDIES, in
ANALYTICAL STUDIES, the investigator proceeds
with a ‘preformed hypo...
From these study designs one can determine:
1. Whether or not a statistical association
exists between a disease and a sus...
CASE-CONTROL STUDY
 Are used to retrospectively determine if there is an
association between an exposure and a specific h...
3 DISTINCT FEATURES OF CASE-
CONTROL STUDY:
1. Both exposure and outcome have occurred before the
start of the study.
2. T...
WHEN IS A CASE-CONTROL STUDY
WARRANTED?
 Conducted before a cohort or an experimental study:
 to identify the possible e...
STUDY DESIGN
9
STUDY
POPULATION
Study begins here
CASES
(disease)
CONTROLS
(no disease)
Factor present
Factor absent
Facto...
 Distinction between 2 studies i.e. case-control and
cohort: not based on the time.
 It is the starting point of the stu...
BASIC STEPS
11
1. Selection of cases
2. Selection of controls
3. Matching
4. Measurement of
exposure
5. Analysis and
inter...
12
Specify the total and the actual
population
•Study population: derived from
the total population
•Total population: may...
 Definition of case
 Inclusion/exclusion criteria
 Eligibility criteria
 Sources of cases
13
SELECTION OF CASES:
SELECTION OF CASES:
Definition of case:
Diagnostic criteria:
 Must be specified before the study is undertaken.
 It shou...
 Incident cases
 newly diagnosed cases
 Prevalent cases
 which have already been diagnosed , a larger number of
cases ...
Risk factors identified using prevalent cases : more related to
survival with the disease than to the development of the
d...
 This can lead to a type of bias.
 What type of bias ??
17
SOURCES OF CASES:
1) Hospital:
 Can be recruited from a hospital, clinic, GP registers.
 Relatively easy and inexpensive...
2)General population:
 Cases of the study disease occurring within a defined
geographic area, during a specified period o...
SELECTION OF CONTROLS:
 Definition of controls
 Inclusion/exclusion criteria
 Sources of controls
 No. of controls
 N...
SELECTION OF CONTROLS:
 Free from disease under study.
 The controls should have undergone the same
diagnostic work up a...
CRITERIA FOR SELECTION OF
CONTROLS:
1. Similar to the cases in all respects other than having
the disease in question ( MA...
SOURCES OF CONTROLS
23
1. Hospital patients
2. Special controls :
1. Relatives
2. Friend
3. Neighborhood controls
4. Gener...
24
Hospital Controls
 Represent sample of ill defined reference population
 Unlikely to be representative of the general...
Relatives
May be unsuitable where genetic conditions are under
study.
Neighborhood controls
Same geographical area ( LOCAL...
HOW MANY CONTROLS ARE NEEDED ?
 CONTROLS OF SAME TYPE :
-minimum no= 1:1
-Maximum no. of controls per case : 4
- Noticeab...
Multiple controls- increase the power of the study
:
1case
4 controls
 MULTIPLE CONTROLS OF DIFFERENT TYPES:-
( Hospital and neighborhood controls, controls with
different diseases)
- Results...
RATIONALE FOR USING TWO
CONTROL GROUPS:
E.g. let us consider the question, “ Did mothers of children
with brain tumours ha...
SOME POSSIBLE RESULTS:
A] Radiation exposure is same in both brain tumor and
other cancer controls, and is higher in in bo...
 Chances of recall bias present ( a well known
epidemiologist; Ernst Wynder, also called it as
“rumination bias” )
- Moth...
B] Radiation exposure in other cancer controls is same as
in normal controls, but is lower than in brain tumor
cases.
32
B...
MATCHING
-Process of selecting the controls so that they are similar
to the cases in certain characteristics, such as age,...
TYPES OF MATCHING:
1. Group matching
2. Individual matching
Group matching( frequency matching):
Proportion of controls wi...
Individual matching (Matched Pairs)
 For each case, a control will be selected
 Controls should be similar to cases in t...
CONFOUNDING
 Concomitant variables
 Associated with both: exposure and the disease
 Distributed unequally in study and ...
EXAMPLE:
37
Consumption of alcohol is a risk factor for oral cancer
History of
alcohol
Oral cancer
Present Absent Total
Pr...
 Conclusion is incorrect: Hidden effect of tobacco use
Risk factor : alcohol use
Confounder: Tobacco use
People who drink...
EXAMPLE TO REMOVE CONFOUNDING
 Stratification: if risk of cancer remains high in both strata –risk is
not due to tobacco ...
CONTROL OF CONFOUNDING
 Identify all Potential Confounding Variables (PCV) right
at the time when research question is be...
MEASUREMENT OF EXPOSURE:
 Obtained by :
 QUESTIONNAIRES
 PAST RECORDS of cases
 INTERVIEWS
 Information about exposur...
2 X 2 CONTINGENCY TABLE
42
a b
c d
Cases
(disease present )
Controls
(disease absent)
Exposed
(risk factor present)
Not ex...
ANALYSIS:
 To Find out:-
1. Exposure rates among cases and controls to suspected
factor.
2. Estimation of disease risk as...
EXPOSURE RATES:
 Exposure rate in cases:-
 Exposure rate in controls:-
44
)( ca
a

)( db
b

a b
c d
D+
E-
E+
D-
(a+c) ...
EXAMPLE:
Case control study of smoking and lung cancer:
Exposure rates:
Cases: = 33/35=94.2%
Controls:
=55/82=67%
45
33
(a...
 Next step is to ascertain whether there is a statistical
association between exposure and the disease.
 To resolve this...
P-value does not imply
causation
47
ESTIMATION OF RISK : ODDS
RATIO
 Measure of strength of association between the risk
factor and outcome.
 The ratio of t...
In case-control study =
Odds of cases being exposed=
Odds of controls being exposed =
Odds ratio=
49
exposedwascontrolatha...
 When is the odds ratio a good estimate of the relative
risk ?
(3 assumptions)
1. Disease under investigation is rare
2. ...
CALCULATING ODDS RATIO IN AN UNMATCHED CASE-
CONTROL
51
Cases
E
E
N
E
N
N
E
E
E
N
Controls
N
E
N
N
E
N
N
E
N
N
6 3
4 7
D+
...
CALCULATING ODDS RATIO IN A MATCHED CASE-
CONTROL
 Concordant pairs : pairs that had the same exposure
 Either both case...
CALCULATING ODDS RATIO IN A MATCHED CASE-
CONTROL
53
Cases
E
E
N
E
N
N
E
E
E
N
Controls
N
E
N
N
E
N
N
E
N
N
a(2) B(1)
C(4)...
 ODDS RATIO :
54
exposedwascontrolathatodds
exposedwascaseathatodds
SIMILARLY,
Here, it is ratio of the number of pairs i...
INTERPRETATION: ODDS RATIO
• =exposure is not related to disease
1
• = risk in exposed is greater than non
exposed (positi...
BIAS IN CASE CONTROL STUDY
 Defined as “any systematic error in the design, conduct
or analysis of a study that results i...
TYPES OF BIAS
57
Selection bias
Self selection
(volunteers bias)
Berkson's bias
Survivorship
(neyman’s) bias
Healthy worke...
SELECTION BIAS
 Error resulting from the way the subjects are selected
Self selection bias/volunteer induced bias
Avoid v...
 Incidence- prevalence bias (Neyman’s bias,survivorship bias)
 Healthy worker effect
Comparison between health status of...
Selection of inappropriate control group
 Controls should be equally at risk of developing the diseases
 Should be selec...
INFORMATION
(MEASUREMENT) BIAS
1. Wrong technique, wrong definitions
2. Recall bias:-
Disease person is more likely to rec...
STRENGTHS OF THE STUDY DESIGN
 Easy to carry out
 Rapid and inexpensive
 Require comparatively few subjects
 Suitable ...
WEAKNESSES OF THE STUDY DESIGN
 Various types of bias may arise:-
 Relies on memory or past records(accuracy may be
unce...
VARIANTS OF CASE CONTROL
STUDY:
1. Nested case control studies
2. Case cohort studies
 These studies are based in a defin...
ADVANTAGES OF EMBEDDING A CASE
CONTROL STUDY IN A DEFINED COHORT
 Problem of recall bias eliminated
 Temporality can be ...
SOME IMPORTANT DISCOVERIES MADE IN CASE
CONTROL STUDIES :
 1950’s : Cigarette smoking and lung cancer
 1970’s: Diethyl s...
CLASSICAL EXAMPLE: 1
Hypothesis : Association of maternal stilbesterol
therapy with tumor (adenocarcinoma)
appearance in y...
 Cases:
1) Seven girls 15-22years of age with adenocarcinoma of vagina ( clear-
cell type ) were taken from the Vincent M...
Data collection: personal interview (standard questionnaire)
Comparison between groups was made regarding 7 risk factors
1...
 Bias in the study:
 Of the candidates for the control group found on
hospital birth lists- 25% could not be located.( s...
CLASSICAL EXAMPLE: 2
Hypothesis : women who too oral contraceptives were at
greater risk of developing thromboembolic dise...
Results:
-Out of 84, 42(50%) of those with venous thrombosis and
pulmonary embolism had been using OCPs, compared
with 14%...
SUMMARY
73
REFERENCES
74
•Park K. Park’s textbook of preventive and social
medicine. 24th ed.
•Gordis L. Epidemiology. 5th ed. Saunde...
75
Prochain SlideShare
Chargement dans…5
×

Case Control Study (ANALYTICAL EPIDEMIOLOGY)

176 vues

Publié le

This presentation will help in better understanding of the topic.

Publié dans : Santé & Médecine
  • Soyez le premier à commenter

Case Control Study (ANALYTICAL EPIDEMIOLOGY)

  1. 1. CASE-CONTROL STUDY Presentation by: Dr. Nidhi Singh Moderator: Dr. Geeta Pardeshi Department of Community Medicine VMMC and Safdarjung Hospital 1
  2. 2. CONTENTS  Introduction  Study Design  Basic steps  Bias in Case Control Study  Strengths and Weaknesses of the study  Variants of case control study  Classical Examples  Summary 2
  3. 3. CLASSIFICATION EPIDEMIOLOGIC METHODS OBSERVATIONAL STUDIES DESCRIPTIVE STUDIES ANALYTICAL STUDIES ECOLOGICAL CROSS- SECTIONAL CASE-CONTROL COHORT EXPERIMENTAL STUDIES RANDOMISED CONTROLLED TRIALS FIELD TRIALS COMMUNITY TRIALS 3
  4. 4. INTRODUCTION  In contrast to DESCRIPTIVE STUDIES, in ANALYTICAL STUDIES, the investigator proceeds with a ‘preformed hypothesis’ regarding a “causal exposure”.  i.e. this ‘particular exposure’ leads to that particular ‘outcome’).  Comparative studies: uses a comparison group 4
  5. 5. From these study designs one can determine: 1. Whether or not a statistical association exists between a disease and a suspected factor 2. If one exists, then what is the strength of association. 5 INTRODUCTION (CONTD)
  6. 6. CASE-CONTROL STUDY  Are used to retrospectively determine if there is an association between an exposure and a specific health outcome.  Backward looking study ( effect to cause study)  Case- reference study  TROHOC STUDY (Reverse of Cohort )  First common approach to test the CAUSAL HYPOTHESIS . 6
  7. 7. 3 DISTINCT FEATURES OF CASE- CONTROL STUDY: 1. Both exposure and outcome have occurred before the start of the study. 2. The study proceeds backwards from effect to cause. 3. Uses control or comparison group to support or refute an inference. 7
  8. 8. WHEN IS A CASE-CONTROL STUDY WARRANTED?  Conducted before a cohort or an experimental study:  to identify the possible etiology of the disease.  Can investigate multiple exposures :  when the real exposure is not known.  When the disease is rare. 8
  9. 9. STUDY DESIGN 9 STUDY POPULATION Study begins here CASES (disease) CONTROLS (no disease) Factor present Factor absent Factor present Factor absent Time Present Past
  10. 10.  Distinction between 2 studies i.e. case-control and cohort: not based on the time.  It is the starting point of the study which decides the type. 10 Exposed Non- exposed Cohort study Diseased No disease Case- control study
  11. 11. BASIC STEPS 11 1. Selection of cases 2. Selection of controls 3. Matching 4. Measurement of exposure 5. Analysis and interpretation
  12. 12. 12 Specify the total and the actual population •Study population: derived from the total population •Total population: may be as broad as mankind or limited to specific groups.
  13. 13.  Definition of case  Inclusion/exclusion criteria  Eligibility criteria  Sources of cases 13 SELECTION OF CASES:
  14. 14. SELECTION OF CASES: Definition of case: Diagnostic criteria:  Must be specified before the study is undertaken.  It should be well formulated and documented.  Should not be altered or changed till the study is over. Eg. Diagnostic criteria formulated by WHO (MI): 1) ECG abnormalities 2)enzyme changes 3)characteristic chest pain Eligibility criteria:  Whether to include incident or prevalent cases? 14
  15. 15.  Incident cases  newly diagnosed cases  Prevalent cases  which have already been diagnosed , a larger number of cases is often available for study. Despite this practical advantage of using prevalent cases, it is generally preferable to use incident cases of the disease in case-control studies of disease etiology. 15
  16. 16. Risk factors identified using prevalent cases : more related to survival with the disease than to the development of the disease (incidence).  E.g.: Prevalent cases - most people who develop the disease die soon after the diagnosis -underrepresented in a study - more likely to include long term survivors. - Cases non representative - any risk factor identified in this group may not be the general characteristic of all the patients with the disease, but only of the survivors 16
  17. 17.  This can lead to a type of bias.  What type of bias ?? 17
  18. 18. SOURCES OF CASES: 1) Hospital:  Can be recruited from a hospital, clinic, GP registers.  Relatively easy and inexpensive to conduct  Select cases from several hospitals in the community, admitted during a specified period of time. -avoids selection bias: Eg. If the hospital from which cases are selected is tertiary care facility, which selectively admits only severely ill patients, any risk factors identified in the study may be risk factors only in persons with severe forms of the disease. 18
  19. 19. 2)General population:  Cases of the study disease occurring within a defined geographic area, during a specified period of time may be included.  Ascertained through – 1. Survey 2. Disease registry 3. Hospital network  Adv :- description of entire picture of the disease in that population  Disadv:-logistic and cost consideration are often high, hence routinely not done. 19
  20. 20. SELECTION OF CONTROLS:  Definition of controls  Inclusion/exclusion criteria  Sources of controls  No. of controls  No. of control groups 20
  21. 21. SELECTION OF CONTROLS:  Free from disease under study.  The controls should have undergone the same diagnostic work up as cases , but have been found to be NEGATIVE.  Equally at risk of developing the diseases  Selected from same population as of cases. 21
  22. 22. CRITERIA FOR SELECTION OF CONTROLS: 1. Similar to the cases in all respects other than having the disease in question ( MATCHING comes into play) 2. Representative of all persons without the disease in the population from which they are selected. Exclusion criteria: patients with diseases known to be associated either positively or negatively with the exposure of interest. 22
  23. 23. SOURCES OF CONTROLS 23 1. Hospital patients 2. Special controls : 1. Relatives 2. Friend 3. Neighborhood controls 4. General population control
  24. 24. 24 Hospital Controls  Represent sample of ill defined reference population  Unlikely to be representative of the general reference population  Differ from people In the community Eg.: prevalence of cigarette smoking is known to be higher in hospitalized patients than in community residents  Adv:-  Readily available  Aware of antecedent exposures better  Willing to cooperate
  25. 25. Relatives May be unsuitable where genetic conditions are under study. Neighborhood controls Same geographical area ( LOCALITY) Eg.: persons working in same factory children attending same school General population controls -from defined geographic area -identify home of a case as starting point , and from there walk past a specified no. of houses in a specified direction , seek the 1st household that contains the eligible control -door to door approach. 25
  26. 26. HOW MANY CONTROLS ARE NEEDED ?  CONTROLS OF SAME TYPE : -minimum no= 1:1 -Maximum no. of controls per case : 4 - Noticeable increase in power is gained only upto a ratio of 1 case to 4 controls. - Multiple controls per case are used to increase the power of the study - For rare diseases potential cases are limited , because the cases cant be increased without extending the study in time to enroll more cases , the option to increase the controls is often chosen 26
  27. 27. Multiple controls- increase the power of the study : 1case 4 controls
  28. 28.  MULTIPLE CONTROLS OF DIFFERENT TYPES:- ( Hospital and neighborhood controls, controls with different diseases) - Results obtained when cases compared with hospital controls will be similar to the results obtained when cases are compared with neighborhood controls. 28
  29. 29. RATIONALE FOR USING TWO CONTROL GROUPS: E.g. let us consider the question, “ Did mothers of children with brain tumours have more prenatal radiation exposure than control mothers?” 2 groups selected for comparison : 1. Normal controls 2. Other cancer controls 29
  30. 30. SOME POSSIBLE RESULTS: A] Radiation exposure is same in both brain tumor and other cancer controls, and is higher in in both compared to normal controls. 30 Brain tumor cases Other cancer controls Normal controls = History of radiation exposure = No history of radiation exposure
  31. 31.  Chances of recall bias present ( a well known epidemiologist; Ernst Wynder, also called it as “rumination bias” ) - Mothers of children with any type of cancer will recall prenatal radiation exposure better than mothers of normal children. 31
  32. 32. B] Radiation exposure in other cancer controls is same as in normal controls, but is lower than in brain tumor cases. 32 Brain tumor cases Other cancer controls Normal controls Use of multiple controls of different types: to take into account possible potential biases. Eg. Recall bias
  33. 33. MATCHING -Process of selecting the controls so that they are similar to the cases in certain characteristics, such as age, sex, socioeconomic status , occupation etc. -Needed to ensure comparability between cases and controls 33
  34. 34. TYPES OF MATCHING: 1. Group matching 2. Individual matching Group matching( frequency matching): Proportion of controls with a certain characteristics is identical to the proportions of cases with same characteristic Eg. 25% cases- married controls will be selected in such a manner that 25%controls are married Prerequisite: all the cases should be selected first 34
  35. 35. Individual matching (Matched Pairs)  For each case, a control will be selected  Controls should be similar to cases in terms of the specific variables or variables of concern. (often used when hospital controls are taken) Matching for universal confounders : Age, Sex 35
  36. 36. CONFOUNDING  Concomitant variables  Associated with both: exposure and the disease  Distributed unequally in study and control groups  Does not lie in the chain of sequence between the exposure and the outcome  Confounder can itself be a risk factor for the disease under study independently  Should be identified before the data is collected 36
  37. 37. EXAMPLE: 37 Consumption of alcohol is a risk factor for oral cancer History of alcohol Oral cancer Present Absent Total Present 80 20 100 Absent 20 80 100 Total 100 100 200 16 20x20 80x80 RATIOODDS Conclusion: risk of getting oral cancer is 16 times higher if person drinks alcohol
  38. 38.  Conclusion is incorrect: Hidden effect of tobacco use Risk factor : alcohol use Confounder: Tobacco use People who drink alcohol are also often the ones who also use tobacco; and tobacco use itself is a direct cause of oral cancer, whether one drinks or not. 38
  39. 39. EXAMPLE TO REMOVE CONFOUNDING  Stratification: if risk of cancer remains high in both strata –risk is not due to tobacco but due to alcohol itself 39 History of alcohol Tobacco users Oral cancer Present Absent Total Present 60 15 75 Absent 20 5 25 Total 80 20 100 Stratum OR= (60x5/15x20)=1 History of alcohol Non Tobacco users Oral cancer Present Absent Total Present 5 20 25 Absent 15 60 75 Total 20 80 100 Stratum OR= (5x60/20x15)=1 After making adjustment for the use of tobacco, the odds ratio in both strata=1 Alcohol by itself has no risk Differential distribution: alcohol and tobacco users: 60/80=75% Few patients who consume alcohol are nontobacco users: 5/20=25%
  40. 40. CONTROL OF CONFOUNDING  Identify all Potential Confounding Variables (PCV) right at the time when research question is being developed.  Control at stages: 1. During planning 2. During analysis 40 During planning During analysis Restriction Stratified analysis Matching Regression analysis
  41. 41. MEASUREMENT OF EXPOSURE:  Obtained by :  QUESTIONNAIRES  PAST RECORDS of cases  INTERVIEWS  Information about exposure should be obtained in precisely the same manner both for cases and controls  This step has susceptibility to various forms of biases. 41
  42. 42. 2 X 2 CONTINGENCY TABLE 42 a b c d Cases (disease present ) Controls (disease absent) Exposed (risk factor present) Not exposed (Risk factor absent )
  43. 43. ANALYSIS:  To Find out:- 1. Exposure rates among cases and controls to suspected factor. 2. Estimation of disease risk associated with exposure (odds ratio) 43
  44. 44. EXPOSURE RATES:  Exposure rate in cases:-  Exposure rate in controls:- 44 )( ca a  )( db b  a b c d D+ E- E+ D- (a+c) (b+d)
  45. 45. EXAMPLE: Case control study of smoking and lung cancer: Exposure rates: Cases: = 33/35=94.2% Controls: =55/82=67% 45 33 (a) 55 (b) 2 (c) 27 (d) Cases (lung cancer) Non smokers Smokers Controls (without lung cancer) )( ca a  )( db b  •Frequency of risk factor : higher in cases than in controls (exposure is associated with the disease)
  46. 46.  Next step is to ascertain whether there is a statistical association between exposure and the disease.  To resolve this: P-value is calculated  Variables under investigation can be: 1. Discrete : Rates, proportions –chi square test( standard error of difference between 2 proportions) 2. Continuous : standard error of difference between 2 means or t-test  If P-value is </= 0.05 :-statistically significant 46 SMALLER THE P-VALUE, THE GREATER THE STATISTICAL SIGNIFICANCE ( Probability that the association is not due to chance alone )
  47. 47. P-value does not imply causation 47
  48. 48. ESTIMATION OF RISK : ODDS RATIO  Measure of strength of association between the risk factor and outcome.  The ratio of the number of ways the event can occur to the no. of ways the event cannot occur. Eg. Probability of winning = 60% Odds of winning = 48 P P   1 ratioOdds 5.1 40 60 60100 60  
  49. 49. In case-control study = Odds of cases being exposed= Odds of controls being exposed = Odds ratio= 49 exposedwascontrolathatodds exposedwascaseathatodds c a d b bc ad
  50. 50.  When is the odds ratio a good estimate of the relative risk ? (3 assumptions) 1. Disease under investigation is rare 2. Cases must be representative of those with the disease 3. Controls must be representative of those without the disease 50
  51. 51. CALCULATING ODDS RATIO IN AN UNMATCHED CASE- CONTROL 51 Cases E E N E N N E E E N Controls N E N N E N N E N N 6 3 4 7 D+ E- E+ D- ODDS RATIO= (a) (b) (d)(c) ad bc = 3.5
  52. 52. CALCULATING ODDS RATIO IN A MATCHED CASE- CONTROL  Concordant pairs : pairs that had the same exposure  Either both cases and controls were exposed, or both were non exposed  Discordant pairs: different exposure (2 combinations)  Cases –exposed Cases - non exposed  Controls – non exposed Controls – exposed  Calculation is based on the discordant pairs only 52
  53. 53. CALCULATING ODDS RATIO IN A MATCHED CASE- CONTROL 53 Cases E E N E N N E E E N Controls N E N N E N N E N N a(2) B(1) C(4) D(2) D+ E- E+ D- Discordant pairs: B and C
  54. 54.  ODDS RATIO : 54 exposedwascontrolathatodds exposedwascaseathatodds SIMILARLY, Here, it is ratio of the number of pairs in which case was exposed and control was not, to the number of pairs in which the control was exposed and the case was not. 4 1 4 b c :RATIOODDS 
  55. 55. INTERPRETATION: ODDS RATIO • =exposure is not related to disease 1 • = risk in exposed is greater than non exposed (positive association)>1 • = risk in exposed is less than non exposed (negative association: possibly protective) <1 55
  56. 56. BIAS IN CASE CONTROL STUDY  Defined as “any systematic error in the design, conduct or analysis of a study that results in a mistaken estimate of an exposure’s effect on the risk of disease.  Can arise in three ways:- 1. Basic measurement technique is wrong 2. Variations between observer or subjects 3. At the time of : 1. Selection 2. Making measurements 56
  57. 57. TYPES OF BIAS 57 Selection bias Self selection (volunteers bias) Berkson's bias Survivorship (neyman’s) bias Healthy worker effect Exposure-related bias Inappropriate group Measurement bias Recall bias Observer bias
  58. 58. SELECTION BIAS  Error resulting from the way the subjects are selected Self selection bias/volunteer induced bias Avoid volunteers: they may be systematically very different from the usual population Berkson’s bias(hospital selective admission) Patients admitted with two concurrent diseases to a hospital may find a stronger association than would really exist in general community 58
  59. 59.  Incidence- prevalence bias (Neyman’s bias,survivorship bias)  Healthy worker effect Comparison between health status of military and civilian population may show a better health status of the soldiers  Exposure related bias: Special type of Berkson’s bias Occurs due to differing probability of hospital admission among those who have and those who do not have the suspected cause 59
  60. 60. Selection of inappropriate control group  Controls should be equally at risk of developing the diseases  Should be selected from same population as of cases. 60
  61. 61. INFORMATION (MEASUREMENT) BIAS 1. Wrong technique, wrong definitions 2. Recall bias:- Disease person is more likely to recall the possible exposure 3. Observer(interviewer’s bias):- Interviewer subconsciously may be more inclined to interrogate /examine the diseased group, to prove the research question 61
  62. 62. STRENGTHS OF THE STUDY DESIGN  Easy to carry out  Rapid and inexpensive  Require comparatively few subjects  Suitable to investigate rare disease  Good for diseases with long latency period  No risk to subjects  Allows the study of several risk factors  No attrition problem  Ethical problems are minimal 62
  63. 63. WEAKNESSES OF THE STUDY DESIGN  Various types of bias may arise:-  Relies on memory or past records(accuracy may be uncertain)  Validation of information obtained is difficult  Selection of an appropriate control group may be difficult  Incidence can not be measured  Do not distinguish between the cause and the associated factors 63
  64. 64. VARIANTS OF CASE CONTROL STUDY: 1. Nested case control studies 2. Case cohort studies  These studies are based in a defined cohort, which is followed over time.  In the beginning baseline data is collected of the cohort  Later on when cases develop, controls are selected accordingly, and then analysis is done only for those cases and controls selected from that population. 64
  65. 65. ADVANTAGES OF EMBEDDING A CASE CONTROL STUDY IN A DEFINED COHORT  Problem of recall bias eliminated  Temporality can be established  More economical study to conduct 65
  66. 66. SOME IMPORTANT DISCOVERIES MADE IN CASE CONTROL STUDIES :  1950’s : Cigarette smoking and lung cancer  1970’s: Diethyl stilbesterol and vaginal adenocarcinoma, post menopausal estrogens and endometrial cancer  1980’s: aspirin and Reyes Syndrome, tampon use and toxic shock syndrome  1990’s : Diet and cancer 66
  67. 67. CLASSICAL EXAMPLE: 1 Hypothesis : Association of maternal stilbesterol therapy with tumor (adenocarcinoma) appearance in young women 67 1) Cancer of vagina: rare disease 2) Usual occurrence in the women above 50 years of age 3) Apparent clustering of cases, which appeared within 4 years (1946-1951)
  68. 68.  Cases: 1) Seven girls 15-22years of age with adenocarcinoma of vagina ( clear- cell type ) were taken from the Vincent Memorial Hospital which occurred between 1966 and 1969 2) An eighth identical case occurred in 20yr old patient, treated in another hospital was also included because she and her family with their matched controls were as available as their own seven cases . Controls: 1) 4 matched controls were selected per case 2) Selected by birth record of the hospital in which each patient was born (socioeconomic differences are reduced) 3) Females born within 5 days and on the same type of service ( ward or private) 68
  69. 69. Data collection: personal interview (standard questionnaire) Comparison between groups was made regarding 7 risk factors 1. Maternal age at the birth 2. Maternal smoking 3. Bleeding during study pregnancy 4. Any pregnancy loss 5. Maternal estrogen therapy during study pregnancy ( DES) 6. Breast feeding of infant 7. Intrauterine x-ray exposure Results: 1) Highly significant association between the maternal estrogen therapy during study pregnancy loss and development of adenocancer of vagina in their daughter (p- value <0.00001) 2) Low level of significance seen with maternal bleeding in the study pregnancy (p-value<0.05) and prior pregnancy loss (p- value< 0.01) 3) No significant difference seen with other factors. 69
  70. 70.  Bias in the study:  Of the candidates for the control group found on hospital birth lists- 25% could not be located.( selection bias) 70
  71. 71. CLASSICAL EXAMPLE: 2 Hypothesis : women who too oral contraceptives were at greater risk of developing thromboembolic disease 1)249 reports of adverse reaction 2)16 reports of death in women taking oral contraceptives Cases: -women admitted to hospitals with venous thrombosis or pulmonary embolism without medical cause Controls: Women admitted to same hospital with other diseases. (2 per case)  Matching done for age, marital status and parity. 71
  72. 72. Results: -Out of 84, 42(50%) of those with venous thrombosis and pulmonary embolism had been using OCPs, compared with 14% of controls - Investigators found that users of oral contraceptives were about 6 times more likely to develop thromboembolic disease. 72
  73. 73. SUMMARY 73
  74. 74. REFERENCES 74 •Park K. Park’s textbook of preventive and social medicine. 24th ed. •Gordis L. Epidemiology. 5th ed. Saunders Elsevier; •Rajiv bhalwar.Textbook of Community Medicine.2nd ed. •Charles H. Hennekens . Epidemiology in Medicine
  75. 75. 75

×