Introduction to Clinical Epidemiology Class 2

1
Introduction to Clinical EpidemiologyClass 2

• Spring 1999 Elective
• UT-H HSC
• Jan Risser, PhD and Will Risser, MD PhD

2
Case Reports / Case Series

• Epidemiology is involved with
• What case definition
• When time
• Where place
• Who person
• Why causes

3
Case Reports / Case Series

• Case reports describe the experience of a single
  patient or group of patients with a similar DX.
• Does halothane cause hepatitis
• rare disease (halothane induced hepatitis)
• single anesthetist with recurrent hepatitis
• symptoms could be elicited with halothane
• hepatitis could be documented
• single case report clarified the
  halothane/hepatitis association

4
Case Reports / Case Series

• Case reports are susceptible to bias.
• NEVER use case reports to make treatment
  decisions.
• CATS (critically appraised topics - PBL)
• avoid cluttering your mind doing a CAT on a case
  series - some are very interesting - but the
  point of a CAT is to help in patient care - and
  more rigorous study designs are necessary.

5
Case Series

• Over a period of 18 months,
• 65 individuals were seen with
• fever, hypotension, diffuse rash, desquamation,
  and impairment of multiple organ systems.
• 59 female, 6 male
• 8 deaths
• age range 8 to 52.

6
Development of Case Definition

• Case definition
• fever, hypotension, rash, and desquamation
• involvement of 3 organ systems
• absence of evidence of other etiologies

7
CDC investigation of Toxic Shock Syndrome

• 1978 - new shock illness described
• staphylococcus aureus infections Todd et al.
• 1979 - 3 new cases reported
• Wisconsin State Health Department - all women
• Surveillance began - by January of 1980 -
• 12 cases, all women identified
• 11/12 menstruating at onset of illness.
• By May of 1980, 55 cases reported / 7 deaths

8
Epidemiologic Approach

• What cases meeting TSS case definition
• When since Todds 1979 report
• Where any where in the U.S. (a bit vague)
• Who women (age, menstruating)
• Why proximate - staphylococcus aureus
• distal - Menstruation?

9
How to evaluate case series

• Is this rare?
• How would we determine this?
• What would we use as denominator?
• Is the association with menstruation real?
• What is the case-fatality rate?
• Is this rate biased?

10
Definitions in Epidemiology

• Bias
• Confounding
• Frequency Measures
• Prevalence
• Incidence
• Measures of Association
• Causal Inference

11
Frequency measures

• Ratio value obtained by dividing one quantity
  by another.
• The ratio of male to female birth in U.S. in
  1979
• 1,791,000 / 1,703,000 1.052
• Proportion a ratio where the numerator is
  always part of denominator
• The proportion of males among all birth in 1979
  
• 1,791,000 / 3,494,000 51.3
• Rate a change in one quantity per unit change
  in another The rate of developing lung cancer
  during a 5 year study is
• 7 / 50 persons5 years or 7 / 250 person-years
• 0.028 cases of lung cancer per person-year

12
Prevalence

• Prevalence (a proportion)
• the proportion of the population at a given time
  that have the factor of interest.
• Prevalence of an exposure
• what proportion of this class have BMI gt 25
• Prevalence of outcome
• what proportion of this class have hypertension
• Point Prevalence - existing cases at a point in
  time
• Period Prevalence - existing cases plus those
  developing over a specified period of time

13
Prevalence

• Numerator
• all those with the attribute at a particular time
• Denominator
• the population at risk of having the attribute
  during that same time period

14
Prevalence

• Choice of denominator may be difficult.
• in 1997 there were 1854 cases of syphilis in
  Harris County
• what should be used for the denominator?
• 55 cases of a new disease reported in three
  states
• what should be used for the denominator?

15
Incidence

• Incidence density the probability (risk) of an
  individual developing the disease (outcome)
  during a specific period of time, using total
  person-time as the denominator. One subject
  followed one year contributes one person-year
  (PY).

16
Incidence

• Cumulative Incidence the probability (risk) of
  an individual developing the disease (outcome)
  during a specific period of time.

17
Incidence, Prevalence
Onset
A
B
C
D
E
F
1994
1986
1988
1990
1992
What was prevalence of disease in 1992? What is
risk of developing disease within 2 years?
1 case (A) / 4 subjects) 25
18
Incidence, Prevalence
Incidence within 2 years
1/6 17
19
Measure of Disease Association

• Ratios
• rate ratio, risk ratio or relative risk (all
  abbreviated RR)
• odds ratio (OR), and
• prevalence ratios.
• Difference measurements of disease frequencies
  include attributable risk

20
Case Reports and Medical Advancement

• These all started with case reports - what study
  design next?
• Lyme Disease (1975)
• Legionellosis (1976)
• AIDS (1981)
• Hantavirus (1993)
• DES exposure (1989)
• EMS l-tryptophan (1970)
• TSS (1980)

21
What next?

• CDC outbreak investigation guidelines
• create case definition
• active case finding
• descriptive epidemiology
• characterize the cases person, place, time
• formulate hypotheses

test hypotheses with case-control studies
22
The case-control study

• Retrospective study design
• identifies cases
• finds controls
• asks about history of exposure
• Measure of association
• odds ratio

23
Retrospective vs. prospective study designs
Retrospective (Case-control)
Disease
Present (Cases)
Absent (Controls)
Risk Factor
a
Present (Exposed)
b
Prospective (Cohort)
d
c
Absent (Not Exposed)
24
Case control studies
Disease Status
Yes
No
Total
b
a
Yes
Exposure Status
a b
No
d
c
c d
b d
a c
N
25
Case-control studies

• Selection of cases
• Case definition is very important
• All cases have an equal probability for
  selection reduce selection bias
• Selection of controls
• Identical in every respect except disease of
  interest

26
Case control studies

• Strengths
• Good for unusual or rare diseases
• Smaller in size, quick, easy, cost-effective
• Can use secondary data on disease
• More easily replicated
• Can test hypotheses

• Weaknesses
• Uncertainty is exposure-disease time relationship
• Representativeness of cases or controls
• Memory problems
• Rare exposure a problem
• Survivor problem
• Bias potential (selection)

27
Case-control and the Odds Ratio
Disease
Y
N
How much risk is too much risk?
a
b
ab
Y
Exposure
cd
d
c
N
ac
N
bd
Odds of exposure if case a / (ac) / c /
(ac) a/c Odds of exposure if control b /
(bd) / d /(bd) b/d Odds exposure given
disease (a/c)/(b/d) (ad)/(cb)
28
Case-control and the Odds Ratio
Y
N
Y
N

29
TSS - 3 case-control studies
(50/0) / (43/7) 17.4
(30/1) / (71/22) 6.4
(12/0) / (32/8) 6.5
NOTE A correction factor of 0.5 was added to
each cell when 1 cell contained 0
30
Study Methods

• CDC - 1 52 TSS cases with age-matched
  acquaintance controls
• Wisconsin Study 31 cases, 93 controls from
  gynecologic clinics, matched only for
  menstruation
• Utah 12 TSS cases, 40 neighborhood-matched
  controls

31
Matched Pair analysis
OR 16 /1 16
How many cases used tampons continually? How many
cases did not use tampons continually? What about
controls?
32
How Big is Big?

• Is an OR of 16 big?
• Is an OR of 16 statistically significant?

33
BRIEF INTERLUDE - STATISTICS

• Before proceeding we need to know a little about
  inference and statistical association

34
How Big is Big?

• Is an OR of 16 big?
• Is an OR of 16 statistically significant?

35

• The whole purpose for doing research is to learn
  something new.
• The result of a research project is the goal
• this is the important information that the
  researchers want the informed public to remember.
• As we read the literature - we should ask
  ourselves
• What is the major result?
• What does this result mean?

36
Statistical Issues in Epidemiology

• We have to remember that epidemiologic studies
  draw inferences about the experiences of an
  entire population based on an evaluation of only
  a sample.

37
Statistical Issues in Epidemiology

• When studying a sample of the population the
  observed associations can be due to

• Bias

• Confounding

• Chance

• Or it can be due to a true association

38
Statistical Issues in Epidemiology

• What do we mean by chance and how does this
  relate to determining a
• true association

• Where do we start?

39
Statistical Issues in Epidemiology

• Association does not mean cause and effect

• Assessing causality involves judgement based on
  the totality of evidence

• Making judgements about causality involves a
  chain of logic that addresses two major areas

1. Whether the observed association is valid
2. Whether the totality of evidence supports a
judgement of causality
40
Statistical Issues in Epidemiology

• The evaluation of the role of chance is done in 2
  steps

1. Estimate the magnitude of the association

• We do this with OR, RR, correlations, AR

2. Hypothesis testing

• Calculate a test statistic, obtain a p value or
  confidence interval

41
Statistical Issues in Epidemiology

• p-value the probability of obtaining a sample
  showing an association of the observed size or
  larger by chance alone under the hypothesis that
  no association exists.

• Confidence interval a range of values that one
  can say, with a specific degree of confidence,
  contains the true population value.

• Sample statistic a number which describes some
  aspect of a sample which represents a population.

42
Statistical Issues in Epidemiology

• This can be done by calculating a test statistic
  of the general format
• The selection of the particular test used depends
  on the specific hypothesis being tested and
  characteristics of the collected data.

43
Statistical Issues in Epidemiology

• If we were to toss a coin 30 times while trying
  to determine if it was a fair coin, and we got 24
  heads, how would we determine if 24 was different
  that the expected number?
• Observed - Expected (under the null)
• Estimated variability in the sample
• We observed 24 - how many did we expect?
• How would we estimate variability?

44
Statistical Issues in Epidemiology

• Observed - Expected (under the null)
• Estimated variability in the sample
• 24/30 - 15/30
• ? Variability
• Variability p(1-p)/n1/2 (24/306/30)/301/2
  0.07
• (24/30) - (15/30) /0.07 4.3
• p lt0.001

45
Statistical Issues in Epidemiology

• The p value indicates the possibility that
  findings at least as extreme as those observed
  were unlikely to have occurred by chance alone.
• In 1000 experiments with 30 tosses with a fair
  coin - we would expect only 1 to result in 24
  heads or more.

46
Statistical Issues in Epidemiology

• A statistically significant finding does not mean
  that the results DID NOT occur by chance - only
  that it is unlikely that they occurred by chance.
  
• A non-significant finding does not mean that the
  results DID occur by chance.

47
Statistical Issues in Epidemiology

• More often in epidemiology we are examining
  discrete data - the 2 x 2 table presents discrete
  data. Here we are testing whether the
  distribution of counts in the 4 cells is
  different than expected under the null
  hypothesis.

48
Statistical Issues in Epidemiology

• But how do we determine the expected value for
  the cells of a 2 x 2 table?
• O Observed Count in a category
• E Expected Count in a category
• å Sum of all categories
• df Degrees of freedom

49
Statistical Issues in Epidemiology

• All tests of statistical significance lead to a
• probability statement
• usually expressed as a p value
• The p-value obtained is based on the principle
  that, given the distribution of interest, it is
  possible to calculate the exact probability or
  likelihood of obtaining a result at least as
  extreme as that observed by chance alone assuming
  there is truly no association.

50
Statistical Issues in Epidemiology

• A probability of 0.05 is the usual (arbitrary)
  cut-off level for statistical significance
• If p lt0.05, we conclude that chance is an
  unlikely explanation for the finding. The null
  hypothesis is rejected, and the statistical
  association is said to be significant.
• If p gt0.05, we conclude that chance cannot be
  excluded as an explanation for the finding we
  fail to reject the null hypothesis.

51
Statistical Issues in Epidemiology

• No p value
• however small - completely excludes chance
• No p value
• however large - completely mandates chance
• p values only evaluate the role of chance
• they say nothing about other alternative
  explanations or about causality
• p values reflect the strength of the association
  and the study sample size

52
Statistical Issues in Epidemiology

• A small difference may achieve statistical
  significance if the sample size is large
• A large difference may not achieve statistical
  significance if the sample size is too small

53
Statistical Issues in Epidemiology

• We address these problems by calculating
  confidence intervals (CI)
• CI indicates the range within which the true
  magnitude of effect lies with a certain degree of
  assurance. The degree of assurance is defined by
  the p value you assign.
• The CI gives all the information of a p value
  PLUS the expected range of effect sizes.

54
Statistical Issues in Epidemiology

• If the null value is included in a 95 confidence
  interval, then the corresponding p value is, by
  definition, greater than 0.05.
• If the null value is not included, the
  association is considered to be statistically
  significant.
• WHAT IS THE NULL VALUE for Odds Ratios and
  Relative Risks (Rate Ratios)?

55
Statistical Issues in Epidemiology

• Test Based CI for either OR or RR
• NOTE variance for either RR or OR may be
  estimated using the chi-square test statistic.
  Miettinen, Am J Epidemiol 103226-235, 1976

56
Statistical Issues in Epidemiology

• Taylor Series to estimate the lnOR variance
  Woolf, Ann Human Gen 19251-253, 1955

Note e is a function on you calculator. You
need a key marked ex and you enter the OR times e
raised to the power of the results between the
brackets .
57
Statistical Issues in Epidemiology

• Taylor Series to estimate the lnRR variance
  Katz, Biometrics, 34469, 1973

Note e is a function on you calculator. You
need a key marked ex and you enter the OR times e
raised to the power of the results between the
brackets .
58
Statistical Issues in Epidemiology

• Inference involves making a generalization about
  a larger group of individuals on the basis of a
  subset or sample.
• The p value indicates the probability or
  likelihood of obtaining a result at least as
  extreme as that observed in a study by chance
  alone, assuming that there is truly no
  association between the study variables.

59
Statistical Issues in Epidemiology

• HOWEVER Before we get to the major result - we
  need to examine several issues
• 1. What was the question that this study
  intended to answer?
• 2. What were the methods used to answer this
  question?
• 3. Are there errors in the study design that
  might invalidate the results?

60
Statistical Issues in Epidemiology

• For the purposes of critical understanding, we
  want to consider information that is often not
  given in the summary.
• Is chance a likely explanation for the results?
• Is selection bias a likely explanation for the
  results?
• Is information bias a likely explanation for the
  results?
• Are the authors conclusions reasonable in terms
  of the information presented?

61
Cancer Case Series
What next with this case-series?
62
DES exposure and Vaginal Cancer

• matched-pair analysis (1 case, 4 controls)
• maternal factors and breast fed
• no statistically significant differences in
  maternal age

63
Matched Case-control

• Case Control OR
• prior pregnancy loss Yes 6 5
• No 2 27
• Estrogens this pregnancy Yes 7 0
• No 1 32
• Breast Feeding Yes 3 3
• No 3 29

64
DES associated with Vaginal Carcinoma

• What are the risks to women exposed to DES
• How could we determine the risks?

65
Cohort studies

• longitudinal or prospective studies
• starts with people free of disease with varying
  degree of exposure from cause to effect
• two points in time, individual is unit of
  observation and analysis