CASECOMPARISON STUDY

1
CASE-COMPARISON STUDY

• A type of observational analytic epidemiologic
  investigation
• Subjects are selected on the basis of whether
  they do (cases) or do not (controls/comparison
  group) have a particular disease under study.
• The groups are then compared with respect to the
  proportion having a history of an exposure or
  characteristic of interest.

2
Purposes include

• Testing a hypothesis about an association between
  a disease and exposure (Example the association
  between lung cancer and passive smoking)
• Exploring data to identify exposures for further
  study data-driven hypotheses (Example the
  association between various exposures and ovarian
  cancer)
• Tracking a point of origin a known infectious
  agent

3
Research Question

• Is there an association between the presence or
  the absence of a particular disease or condition
  and one or more specific attributes
  (exposures/risk factors)?
• OR
• Among those with a disease is an attribute more
  or less prevalent than among those without the
  disease?

4
Distinguishing Characteristics

• Looks for an association by comparing a group of
  diseased individuals to a group of non-diseased
  individuals
• offers a solution to the difficulties of studying
  diseases with long latency periods (DES and
  vaginal cancer)
• Particularly efficient in terms of both time and
  costs relative to other analytic designs
• Particularly well suited to the evaluation of
  rare diseases which would otherwise need to
  follow a large number of individuals in order to
  accumulate a significant number of who develop a
  particular outcome

5
Potential problems

• Both exposure and disease have already occurred
  at the time participants enter into the study
• Particularly susceptible to bias
• Requires careful consideration of the sources of
  bias which may arise to minimize its occurrence

6
Issues in design of case-comparison studies

• Definition of cases
• Case definition should be as homogeneous as
  possible.
• Strict diagnostic criteria needs to be
  established
• Selection of Cases
• Hospital or medical care facilities
  (hospital-based case comparison study) during a
  specified period of time. Relatively easy and
  inexpensive
• Defined population (population-based
  case-comparison study)- select all affected
  individuals from a population or a sample of
  affected individuals. Avoids selection factors
  that lead an affected individual to utilize a
  particular health care facility or physician

7
Issues in design of case-comparison studies

• Incident cases (newly diagnosed cases)
• Prevalent cases (existing cases at a point in
  time)
• Can increase sample size but can reflect
  determinants of duration and the development of
  the disease
• More difficult to ensure reported exposures
  relate to time before the development of the
  disease

8
Issues in design of case-comparison studies

• Selection the comparison group is the most
  difficult but critical issue of this design
• The comparison group is necessary to allow the
  evaluation of whether a frequency of an exposure
  or specified characteristic observed in the case
  group is different from that which would have
  been expected based on the experience of a series
  of comparable individuals who do not have the
  disease
• Involves consideration of issues
• characteristic and source of cases
• the need to obtain comparable information from
  cases and controls
• practical and economic issues

9
Issues in design of case-comparison studies

• The comparison group should be selected not to
  represent the entire non-diseased population but
  the population of individuals who would have been
  identified and analogous to selection of cases
• The comparison group should be comparable to the
  source population of the cases and that any
  exclusions or restrictions made in the
  identification of cases apply equally to the
  comparison group and vise versa

10
Issues in design of case-comparison studies

• Sources of comparison group
• hospital- subjects with conditions other than the
  disease under study
• Easy to identify and readily available thus
  reducing cost and effort
• More likely than healthy individuals to be aware
  of antecedent exposure or events (reduce
  potential for recall bias)
• More likely to be comparable to cases on those
  factors that influence cases to have a particular
  health provider (SES, residency)
• May differ from healthy individual in ways that
  may be associated with the disease or likelihood
  of hospitalization therefore not representing the
  distribution of exposure in the population from
  which cases were derived. Studies have shown that
  hospitalized individuals are more likely to
  smoke, use oral contraceptives, and be heavier
  drinkers)
• Important to consider which category of diseases
  should be included for comparison group and that
  these diseases are not associated with the risk
  factor under study

11
Sources of comparison group

• general population
• households in the targeted neighborhoods
• random digit dialing
• population registries (i.e., drivers license,
  voting lists)
• Advantages include healthy population that may
  better represent source of cases
• Disadvantages include increased time, costs,
  differences in quality of the information, lower
  participation rates, participants may differ from
  non-participants.

12
Sources of comparison group

• special groups
• friends
• relatives
• spouses
• twins
• neighbors
• Advantages are that they are healthy and more
  likely to be cooperative. Provides control of
  potential confounding factors (SES, ethnicity,
  and environmental exposures)
• Disadvantage is that if the study factor itself
  is very similar to cases, may underestimate the
  true effect of the exposure of interest.
• More expensive
• More time consuming
• Cases unwilling to get friends to fill this role

13
Issues in design of case-comparison studies

• Number of comparison groups
• When there is concern about comparability of
  comparison group, it may be advantages to use
  several control groups
• Number of comparison subjects
• when sample size is large, the optimal ratio for
  case and comparison subjects is 11
• when sample size is limited, with small numbers
  being available or when costs of obtaining
  information is greater for cases than for
  controls, the case to comparison group ratio can
  be altered to achieve the desired sample size.
• As the number of comparison subjects per case
  increases, the power of the study increases
  however, beyond 4 comparison subjects per case,
  there is very little benefit

14
Issues in ascertainment of exposure status

• Sources of exposure information should be
  carefully considered in terms of the ability to
  provide accurate and comparable information for
  all study groups
• Information can be obtained from
• interviews
• questionnaires
• recorded information
• Information can be obtained from study subjects
  themselves or from surrogates (spouse, siblings,
  friend, parent, co-workers, etc)
• Interviewer or abstracter should be blinded as
  much as possible to case status and hypotheses
  being tested

15
Ascertainment of exposure (continued)

• Information should be obtained under similar
  circumstances for all subjects (methods of
  obtaining and place)
• Ideal to obtain information recorded prior to
  disease (e.g., X-ray exposure, birth certificate
  information, work records, medical records, etc.)
• Key issue is how to define exposure
• What part of the person's history should be
  considered exposed?

16
The Analysis

• The formal hypothesis under investigation may be
  stated as follows
• If the disease and the attribute (purported
  exposure or risk factor) are not associated, then
  the measure of risk or association will equal
  1.0
• OR stated differently
• If the prevalence of the exposure is the same
  among the disease (case) and non-diseased
  (comparison) subjects, then the measure of risk
  or association will be equal to 1.0
• The appropriate measure of association in a
  case-comparison study is the ODDS RATIO (OR). It
  is defined as the ratio of two sets of odds
• 1. The odds of being exposed to non-exposed among
  cases
• 2. The odds of being exposed to non-exposed among
  controls

17
The Analysis

• Odds provides the same information as prevalence,
  except in a different form. Recall the numerator
  of prevalence is a subset of the denominator,
• Example The prevalence of being African American
  is 13. Out of every 100 individuals, 13 are AA.
• OR
• The probability that a member of the population
  is AA is 13
• Odds are a ratio where the numerator and
  denominator are mutually exclusive
• Example The odds of being African American to
  races other than AA is 13/87
• Odds can be converted to prevalence and
  prevalence can be converted to odds

18
The analysis

• At the beginning of a case comparison study the
  2X2 table appears as

19
The analysis

• At the end of a case comparison study the 2X2
  table appears as

The relationship between the size of the diseased
and non-diseased groups is not determined by the
natural course of the disease, but is set by the
investigator based on statistical considerations
20

• The odds ratio is a close approximation of the
  relative risk when the disease being studied is
  relatively rare
• The odds ratio is the appropriate calculation to
  test the association of a case-comparison study.
• It is not and should not be mistaken for a
  relative risk

21
Example of Odds Ratio
The odds of being exposed among diseased is
A/C The odds of exposure among nondiseased is
B/D The ratio of these Odds is A/C or AD
B/D CB
22
Example of Odds Ratio HPV SIL
The odds of being exposed among diseased is
A/C The odds of exposure among nondiseased is
B/D The ratio of these Odds is A/C or AD
B/D CB
23
Example of Odds Ratio HPV SIL
The odds of being exposed among diseased is
248/75 The odds of exposure among nondiseased is
54/216 The ratio of these Odds is 3.31 or
53568
0.25 4050 OR 13.22
24
Confidence Intervals

• Confidence intervals (CI) represent the range
  within which the true magnitude of the effect
  lies with a certain degree of assurance
• CIs provide all the information of the P value in
  terms of whether the association is statistically
  significant at a specified level.
• If the null value (OR1.0) is included in the 95
  CI, then the corresponding P value by definition
  is greater than .05

25
Confidence Intervals

• If the null value is not included in the
  interval, the the corresponding P-value is less
  than .05 and the association is statistically
  significant.
• The width of the CI indicates the amount of
  variability inherent in the estimate and thus the
  effect of the sample size.
• The larger the study, the more stable the
  estimate, the narrower the CI
• The wider the CI, the greater variability in the
  estimate of effect, and the smaller the sample
  size

26
Confidence Intervals

• Construction of the Confidence interval for the
  Odds ratio
• Takes into account the distribution is highly
  skewed to right (e.g., RR can never be lt 0, but
  can assume a positive value up to infinity)
• If we apply a transformation to the the RR-
  specifically taking the natural logarithm,
  (denoted ln) will result in an approximately
  normal distribution
• Therefore, we construct the upper and lower
  bounds of the CI around ln(RR). Then getting back
  to the RR, we take the antilogarithm of the
  ln(RR). (written as EXPln RR or
• e ln RR

27
Confidence Intervals

• The general formula (Taylor series) is
• Where z is the value of the standard normal
  distribution associated with the desired level of
  confidence

28
Confidence Intervals Taylor series
CI 8.92, 19.60
29
The Interpretation

• The general interpretation of the OR is as
  follows
• Diseased individuals have OR times the odds of
  exposure than non-diseased individuals
• Often used is
• Exposed individuals are OR times more likely to
  have the disease than non-exposed
• Or a more liberal interpretation
• Exposed individuals are OR more likely to develop
  the disease than non-exposed indiviudals

30
Case comparison studies referent group

• The referent group always has an odds ratio of
  1.0
• How is the referent group selected?
• May be arbitrary, but should reflect the
  hypothesis under investigation
• When calculating the odds ratio using multiple
  categories, one category must be selected as the
  unexposed group (the unexposed group)

31
Odds ratio for the association between smoking
and MI
32
Odds ratio for the association between smoking
and MI
33
Odds ratio for the association between vitamin C
intake and oral cancer
34
Odds ratio for the association between vitamin C
intake and oral cancer
35
ABSTRACT Objectives. This study explored the
risk of childhood acute lymphoblastic leukemia
(ALL) associated with participation by household
members in hobbies or other home projects
involving organic solvents.Methods. Participants
in this case-control study were 640 subjects with
ALL and 640 matched controls.Results. Childhood
ALL was associated with frequent (gt4 times/month)
exposure to model building (odds ratio OR
1.9 95 confidence interval 95 CI 0.7, 5.8)
and artwork using solvents (OR 4.1 95 CI
1.1, 15.1). We also found elevated risk (OR
1.7 95 CI 1.1, 2.7) among children whose
mothers lived in homes painted extensively (gt4
rooms) in the year before the children's
birth.Conclusions. In this exploratory study,
substantial participation by household members in
some common household activities that involve
organic solvents was associated with elevated
risks of childhood ALL
American Journal of Public Health
Volume 91(4)             April 2001           
 pp 564-567
Household Solvent Exposures and Childhood Acute
Lymphoblastic Leukemia
Freedman, D. Michal PhD et al