Designing a New Study:

1
Designing a New Study

• II. Cross-sectional and Case-control Studies

2
A cohort study the sequence of making the
measurements is the same as the chronology of
cause and effect.

• Cross-sectional (prevalence) study
• All measurements on a single occasion.
• Determine predictor and and outcome after the
  data collection
• Estimate prevalence

• Case-control study
• Begin with the outcome, then identify the
  predictor
• Explore the potential association

3
CROSS-SECTIONAL STUDIES

• Well suited to the goal of describing variables
  and their distribution patterns
• Structure
• Similar to cohort study (except that measurements
  are made at once)
• Can examine associations based on the
  investigators hypothesis, not based on the study
  design.
• e.g., age, race---usually predictors
  blood lead level and hyperactivity
  ----gtmisleading (historic information on
  the time course)

4
Cross-sectional Study
Risk factor No disease
Risk factor Disease
Population
No risk factor No Disease
No risk factor Disease
Sample
Steps 1. Select a sampling from the
population 2. Measure predictor and outcome
variables
5
CROSS-SECTIONAL STUDIES

• Designing
• Settle on the research question
• Specify criteria for the target and accessible
  populations
• Establish the design for drawing the sample
• Decide the phenomena to study
• Define the approach to measuring appropriate
  variables.Example 8.1 Sexually transmitted
  disease and the use of oral contraceptives

6
Statistics for expressing disease frequency in
observational studies
Type of Study Statistics
Definition
of people who have the disease at one point in
time
Cross-sectional Prevalence Cohort
Incidence
of people at risk at that point
of new cases of disease over a period of time
of people at risk during that period
Relative prevalence Relative risk
prevalence/incidence bias
7
Cross-sectional Studies

• weakness
• Difficult to establish causal relationship
• Impractical for the study of rare diseases or
  risk factors from a general population
• e.g., 1 in 10,000 in a general population for a
  stomach cancer in 45-59 year old men
• Susceptible to prevalence/incidence bias
• e.g., Kids with HLA-A2 were at increased risk
  factor for the incidence of leukemia ???truth
  was that HLA-A2 kids live longer

• strength
• Relatively fast and inexpensive
• No waiting time to see the outcome
• No loss to follow-up
• Provide the prevalence of a disease or a risk
  factor
• Convenient for examining networks of causal links
• alcohol intake and HDL-cholesterol
• First step for investigations

8
Cross-sectional Studies

• Case series cross sectional studies for
  relatively rare diseases
• the sample from a diseased population not from a
  general population
• suitable to describe the characteristics of the
  disease than to analyzing differences between
  these patients and healthy people

e.g., Of the first 1000 patients with AIDS, for
example, 727 were homosexual or bisexual males
and 236 were I.V. drug abusers. Furthermore,
within a sample of patients with a disease there
may be association of interest, the higher risk
of Kaposi sarcoma among AIDS patients who are
homosexual than among AIDS patients who are I.V.
drug abusers.
9
Cross-sectional Studies

• Serial Surveys
• To draw inferences about changing patterns over
  time.
• e.g., census data
• Not a cohort study (i.e., following a single
  group of people)

10
Case-Control Studies

• Structure
• the prevalence of the risk factor in subjects
  with the disease (cases) can be compared with the
  prevalence in subjects without the disease
  (controls).
• Retrospective
• house redmore modest and a little riskier than
  the other selections, but much less expensive and
  sometimes surprisingly good!

11
Case-control design
Population with disease (cases)
THE PAST OR PRESENT
THE PRESENT
Risk factor present
Risk factor absent
Disease
Sample of cases
Risk factor present
Risk factor absent
No disease
Sample of controls
Much larger population without disease (controls)
12
Case-Control Studies

• Steps
• 1. Select a sample from a population of people
  with the disease (cases)
• 2. Select a sample from a population at risk that
  is free of the disease (controls)
• 3. Measure predictor variables

13
Designing a case-control study

• Settle on the research question
• Specify criteria for the target and accessible
  populations (the cases and the controls)
• Establish the design for drawing the sample
• Decide the phenomena to study
• Define the variables and measurement approaches,
  and establishes the hypotheses to be tested.

14
Designing a case-control study

• Settle on the research questionWhether there
  is an association between use of aspirin and the
  development of Reyes syndrome

15
Designing a case-control study

• Specify criteria for the target and accessible
  populations (the cases and the controls)
• The cases Children with a viral infection
  followed by Reyes syndrome
• The controls Children with a viral infection
  but no Reyes syndrome

16
Designing a case-control study

• Establish the design for drawing the
  sample(cases)all 30 patients with Reyes
  syndrome who are accessible to him for study
• Establish the design for drawing the
  sample(controls)60 patients drawn from the
  much larger population of accessible patients who
  have had minor viral illnesses without Reyes
  syndrome

17
Designing a case-control study

• Decide the phenomena to study
• Define the variables and measurement approaches,
  and establishes the hypotheses to be tested.ask
  the subjects in both groups about their use of
  aspirin.approximate relative risk can be
  computed

18
Designing a case-control study

• Cannot yield estimates of the incidence or
  prevalence of a disease, because the proportion
  of study subjects who have the disease is
  determined by how many cases and how many
  controls the investigator chooses to sample,
  rather than by their proportions in the
  population.
• Can yield some descriptive information on the
  characteristics of the cases and an estimate of
  the strength of the association between each
  predictor variable and the presence or absence of
  the disease (odds ratio).

19
Odds ratio and relative risk
Disease No disease
Risk factor present Risk factor absent
a b c d

• Odds ratio in a cross sectional studiesad/cb

a/b
ad/cd
c/d
.
a/(a b)
c/(c d)
.
a ( c d)

c (a b)
20
Case-Control Studies

• Strengths
• Efficiency for rare outcomes
• high yield of information from relatively few
  subjects
• Usefulness for generating hypotheses

• Weaknesses
• no incidence
• no prevalence
• no attributable or excess risk
• Sampling bias, and how to control it
• randomization is near impossible (pts with a
  diagnosed)
• misdiagnosed or misdiagnosed are omitted

21
Case-Control Studies

• Weaknesses
• selection of cases ---relatively
  straightforward
• selection of controls---??
• Sampling the cases and controls in the same way
• Matching
• Using two or more control groups
• Using a population-based sample

• Differential measurement bias, and how to control
  it
• Use of data recorded before the outcome occurred
• Blinding