Title: Community Health
1Community Health Epidemiology LMCC 2008 Review
- Part I Biostatistics Epidemiological Methods
- Part II Non-infectious Disease Epidemiology
- Part III Infectious Disease Epidemiology
- Dr. Allen Ross, M.D., Ph.D.
- Associate Professor
- Community Health
- Epidemiology
2Part I Biostatistics Epidemiological Methods
3Descriptive Statistics
4Biological Variability
- All biological measurements are subject to
variability - It is important to be able to describe the
occurrence of a biological factor e.g. height in
a population - The mean, median and mode are the three most
important indices of central tendency - For any symmetrical distribution mean, median and
mode are equal
5Mean
- Mean sum of the observed measurements
- number of observations
- The arithmetic average
6Median
- Median that measurement below which half the
measurements fall, the 50th percentile - The length of hospital stay for nine patients
- 1, 1, 3, 4, 8, 9, 12, 13, 15
- the median is the middle number 8
7Mode
- The most frequently occurring observation. If
more than one value occurs frequently the
distribution can be bimodal or multi-modal - e.g. for values 1,4,3,1,2 the mode is 1
- for values 2,4,2,3,1,5,1 the distribution is
bimodal as 1 2 occur most often
8- Example
- To determine the average length of stay for six
patients undergoing cholecystectomy. The length
of stay in days for each patient is 1, 3, 2, 2,
4, 5 - Qes. Calculate mean, median and mode
- a) Mean 122345 17/6 2.83
- 6
- b) Median 2.5 (the average of the 2 middle
numbers for an even number of observations) - c) Mode 2
-
9The Normal Range
- Sometimes it is necessary to calculate a normal
range e.g for hemoglobin levels, weigh-height
charts - This is usually done by calculating the mean /-
2 standard deviations 95 under the curve - Anything falling outside this is deemed abnormal
10Problems with The Normal Range
- If the data is skewed these rules do not apply
- In medicine what is normal in the statistical
sense may be abnormal clinically - It is better to use a cut off that is clinically
significant
11Strength of Association
- Large relative risk or odds ratio
- Statistically significant (p value lt 0.05)
12Hypothesis Testing
- Is the risk of leukemia associated with maternal
irradiation in utero? - Is the risk of a certain illness associated with
a particular drug treatment in a previous illness - Is the risk of death from lung cancer associated
with smoking?
13Hypothesis Testing
- The null hypothesis states there is no
relationship between exposure and disease. RR 1
or OR 1 - The alternative hypothesis states that there is a
relationship between the exposure and the
disease. RR 1 or OR 1
14Type l Error (?)
- Stating that there is an effect when there really
is not - To mistakenly accept the experimental hypothesis
and reject the null hypothesis - ? is the probability of making a type 1 error and
is p (usually lt.05) - p is the probability of making a type 1 error
15Type ll Error (?)
- Stating that there is not an effect or difference
when there really is - To fail to reject the null hypothesis when in
fact H0 is false - ? is the probability of making a
- type ll error
- 1- ? is also the power of a study
16Testing for Significance using P-values
- The P - value indicates the likelihood of
obtaining a result at least as extreme as that in
the study by chance alone - P value is set at lt 0.05 for medical research
- Dependent on the sample size
- Should not use P-value as a substitute for common
sense - Always determine if the association is clinically
significant as well as statistically significant
17Confidence Intervals
- Provides an interval range around the the odds
ratio or the relative risk - Represents the range within which the true
magnitude of effect lies - Usually set at the 95 level equivalent to Plt
0.05 - Provides all the information of the P value
- If the interval does not contain 1 then the
association between the variables is significant
18- Example 1
- A study is designed to investigate the
association between body fat and breast cancer.
The results show a risk ratio of 6.0, however the
95 confidence interval is (0.8 - 23.2). - Since the interval includes 1, the results may be
due to chance alone and the null hypothesis is
accepted
19- Example 2
- A study is designed to investigate the
association between fluoride and decreased dental
caries. The results show a risk ratio of 3.3 (95
CI 2.2-4.0). - Since the interval does not include 1, the null
hypothesis is rejected and the result is found to
be significant
20- 95 CI (OR, RR) does not contain 1
- P lt 0.05
- Significant finding
- Reject the null hypothesis
- Outcome not do to chance
- 95 CI (OR, RR) contains 1
- P gt 0.05
- Not a significant finding
- Accept the null hypothesis
- Outcome may be do to chance
21Question 1.
- The mean birth weight of first-born
infants of 23 women who smoked more than one pack
of cigarettes per day during pregnancy was 200 g
lower than that of the first-born infants of 16
women who never smoked. The difference was
statistically significant at the 5 level (P lt
0.05). This means which of the following? - A) Smoking during pregnancy retards fetal growth
- B) The difference observed between mean birth
weights was too large to have occurred by chance
alone - C) The difference observed between mean birth
weights could have easily occurred by chance
alone - D) The number of patients studied was not
sufficient to achieve a conclusive result - E) Smoking during pregnancy does not influence
fetal growth
22Question 1.
- B) The difference observed between mean birth
weights was too large to have occurred by chance
alone
23Question 2.
- Fifty known diabetics, all on insulin
therapy, were compared with 50 non-diabetics. The
diabetics showed a higher proportion of neurotic
responses to a questionnaire (P lt 0.05). Which of
the following could be ruled out as a viable
explanation for this finding? - A) Insulin therapy
- B) Chance
- C) Age
- D) Diet
- E) Medical complications of diabetes
24Question 2.
25Question 3.
- As part of a routine physical examination,
uric acid was measured for a 35 year-old male and
found to be 7.8 mg/dl. The normal range for
uric acid for that laboratory is 3.4 to 7.5
mg/dl. If this individual is asymptomatic, which
of the following is a viable explanation? - A) His level is within 2 standard deviations of
the mean for healthy individuals - B) His level is below the 97.5th percentile for
health individuals - C) He is among the small proportion of healthy
individuals who yield high serum uric acid
readings on a given test - D) The normal range was derived from a positively
skewed distribution - E) The normal range was derived from a negatively
skewed distribution
26Question 3.
- C) He is among the small proportion of healthy
individuals who yield high serum uric acid
readings on a given test
27Question 4.
- The following regression equation was
developed from a study of 16 newly diagnosed
diabetics who received phenformin for a period of
one year - L -34
0.29 W - where L is the patients weight loss one
year after therapy began and W is the patients
initial weight. Which of the following could be
deduced from this information? - A) All patients last at least 34 pounds during
the first year of therapy - B) The regression line has a positive slope
- C) The correlation is very strong
- D) Patients who weighed more that others at the
beginning of therapy lost more weight, on the
average, during the first year of therapy - E) The regression line has negative slope
28Question 4.
- B) The regression line has a positive slope
29Vital Statistics, Rates, Standardisations
30Incidence
- One of the most important rates in epidemiology
- Measures the rate at which people without a
disease develop the disease during a specific
period of time - One of two common ways of comparing frequency of
disease in populations (the second is prevalence) - A measure of risk
- Can be described as cumulative incidence or
incidence density
31Cumulative Incidence
- New cases occurring in a given period x 10n
- Population at risk during the same time period
- Incidence Density
- New cases of a disease in a given period x 10n
- Total person-time of observation
32Oral contraceptive and bacteriuria in a community
based studyNEJM 299536, 1978
- Population 2390 women aged 16-49 years who were
free from bacteriuria - 482 were OC users in 1973
- Second survey 1976 showed that 27 had developed
bacteriuria - Cumulative incidence 27 / 482 5.6
33Calculation of Person-years for Incidence Density
Cases Total time Subject A
------------------- 2
years Subject B -------------X
1 years Subject C
----------------------- 2 years Subject D
----------------------------------X 3
years x developed disease -- time
followed Incidence Density 2 / 8
25 per 100 person-years
34A Prospective Study of post-menopausal hormones
and coronary heart diseaseNEJM 3131044, 1985
Population 32,317 postmenopausal women Cases of
coronary heart disease 90 Time period
105,786.2 person-years Incidence density 90 /
105,786.2 person-years
85.1 / 105 person-years
35Issues in Calculation of Incidence
- The denominator should not include those not at
risk but it is often impossible to exclude
persons based on risk - e.g. the incidence of endometrial cancer should
be calculated excluding women s/p hysterectomy - Incidence of endometrial cancer 1960-1973 was
underestimated by 45
36Prevalence Rate
- The second most common measure of disease
frequency - The proportion of persons in the population who
have a particular disease at a specific point in
time (point prevalence) or over a specified
period of time (period prevalence) - All cases during a given time period x 10n
- Population at risk during the same time period
37Point vs. Period Prevalence
- Point Prevalence examines prevalence at a single
point in time - Period Prevalence examines prevalence over a
longer period e.g. a year
38Framingham Heart Study Prevalence Study to
determine the rate of cataract in the population
Population individuals 52 to 85 years of age
2477 persons examined 310 individuals with
cataract Prevalence number with cataract
total population
310/ 2477 12.5
39Risk Ratio / Relative Risk
- RR estimates the magnitude of an association b/w
exposure and disease and indicates the likelihood
of developing the disease in the exposed vs.
non-exposed - RR of 1 indicates identical risk in the two
groups gt1 increased risk lt 1 decreased risk - RR used in cohort studies not case-control
studies
40The following table demonstrates the rates of
pellagra in two groups, males and females
- Risk of illness among females a/ (ab)
46/1484 .031 - Risk of illness among males c/ (cd) 18/1419
.013 - Risk Ratio Ie / Io .031 / .013 2.4
- The risk of pellagra in females is 2.4 times
higher than the risk in males
41Attributable Risk
- The absolute effect of the exposure or the excess
risk of disease in those exposed vs. not exposed - AR Ie - Io
- Ie is incidence in the exposed
- Io is incidence in un-exposed
42Attributable Proportion
- The measure of the public health impact of a
causative factor - Also called the attributable risk percent
- AR / Ie
- Risk for the exposed group - Risk for the
unexposed group x 100 - Risk for the exposed group
43Death Rates and Rate Ratios from lung cancer by
daily cigarette consumption ( Doll and Hill
1951-1961)
44Calculate attributable proportion in smokers of
1-14 cigarettes/day Identify the exposed group
rate 0.57 per 1000/ year Identify the unexposed
group rate 0.07 per 1000 / year calculate the
attributable proportion 0.57- 0.07 x 100
0.877 x 100 87.7 0.57 Therefore 88 of
lung cancers in smokers of 1-14 cigarettes/day
are attributable to smoking
45Mortality Frequency Measures
- A mortality rate is a measure of the frequency of
occurrence of death in a defined population
during a specified interval - Mortality Rate
- Deaths occurring in a given time period x 10n
- Size of the population among which the deaths
occurred
46Mortality Rates
Deaths occurring in a given time period x
10n Size of the population among which the deaths
occurred
- The denominator is usually the midpoint
population - There are multiple measures of mortality such as
crude death rates, cause specific death rates,
infant mortality rates, death to case ratio
47Infant Mortality Rates
- The number of deaths in the first year after live
birth about 10 million infants per year
world-wide - Globally the rates range from 5-6 deaths/1000
live births to 150 deaths/live births - High infant mortality is strongly linked to low
birth weight (lt2500g) in rich countries - In poor countries the association is complex
other associations include GDP,
education/literacy, availability of healthcare
e.g facilities and personnel
48Calculation of Infant Mortality Rates
- Example
- In 1988, 38,910 infants died and 3.9 million
children were born. - The IMR (number of deaths among children less
than 1 year old) - 38910 / 3.9 million
- 9.98 per 1000
49IMR
- Japan, Sweden, Finland 5/1000 live births
- UK 8/1000 live
births - Afghanistan 150/1000 live births
- Canada 5/1000 live births
- African-Americans are twice as likely to die as
white infants
50Mortality Rates
- Crude mortality rates reflect both specific
mortality rates and population composition - Age-specific rates
- Age-adjusted or standardized mortality (a
weighted mortality rate that accounts for
population composition) is used when comparing
regions or countries
51Adjusted Mortality Rates
- Because Crude mortality rates reflect both
specific mortality rates and population
composition, 2 populations can only be compared
when using age specific mortality rates and
age-adjusted rates
52Age-Specific Mortality Rates
- Age is a continuous variable that has a profound
effect on mortality rates - The mortality rates are listed for each age
category. This allows direct comparison between
states or countries - Florida has a higher crude mortality rate than
Alaska, yet when using age-specific rates the
rates of death are noted to be similar - It is always advisable to look at the mortality
rates closely are they crude, age-specific or
adjusted?
53Age-Adjusted Mortality Rates
- It can be tedious to compare every age-specific
rate across towns, countries etc. - The age-adjusted rate is a mathematical way to
apply the age-specific rates to a standardized
population with a fixed numbers in each strata.
This allows easy comparison of rates.
54Cause-Specific Mortality
- Mortality rates for any specific disease may be
stated for the entire population or any subgroup - Deaths assigned to the specified disease per
year x 105 - Population at mid-year
- Crude 1980 cancer mortality rate 416,481 /
226,546,000 yr -
183.8/105/year
55Case-Fatality Rate
- The probability of death among diagnosed cases
- Deaths assigned to the disease in a certain year
x 100 - Total cases of that illness in the same year
- e.g. In 1981-82 200 cases of Reyes syndrome were
reported in the USA for individuals lt18 y and 70
died. Thus, CFR 70 / 200 x 100 35
56Proportionate Mortality Ratio (PMR)
- Proportion of the overall mortality that may be
ascribed to a specific cause. - PMR Deaths assigned to the disease in a certain
year x 100 - Total deaths in the population in the same
year - e.g. PMR of heart disease was 37 in 1985 in the
USA vs. 22 for cancer
57Question 5.
- A study covering records of 150 consecutive
unselected female patients with
hyperparathyroidism at a hospital revealed that
43 were under 45 years old and 107 were 45 years
or older. The author concluded from these data
that in women, the incidence of
hyperparathyroidism is higher in the menopausal
and postmenopausal age groups than it is in the
pre-menopausal age groups. - (a) this conclusion is correct
- (b) this conclusion is incorrect because the
comparison was not based on proper rates - (c) this conclusion is incorrect because there
was improper interpretation of statistical
significance - (d) this conclusion is correct if the author had
compared like with like - (e) this conclusion is incorrect because
observer bias may account for the results
58Question 5.
- (b) this conclusion is incorrect because the
comparison was not based on proper rates
59Question 6.
- Suppose that you are investigating the
possible association between cigarette smoking
and cancer of the lung, and you obtain the
following rates of death from lung cancer among
Canadian males age 35 and over, related to
smoking habits - cigarette smokers - 2.0 per 1,000 per year
- non-smokers - 0.2 per 1,000 per year
- Based on the above observation, which of the
following statements is correct? - (a) the relative risk is 1.8 per 1,000 per year,
and the attributable risk is 10.0 - (b) the relative risk is 10.0 and the
attributable risk is 2.2 per 1,000 per year - (c) the relative risk is 10.0 and the
attributable risk is 1.8 per 1,000 per year - (d) the relative risk is 0.1 and the
attributable risk is 1.8 per 1,000 per year - (e) none of the above
60Question 6.
- (c) the relative risk is 10.0 and the
attributable risk is 1.8 per 1,000 per year
61Question 7.
- The following data were obtained from a
study in which 200 renal cancer cases and 200
controls were interviewed on their previous
employment in a certain industry. -
-
Renal Cancer Cases Controls - Previously employed in this industry
22 20 - No previous employment in this industry
178 180 - Total
200 200
- The observed difference in proportion in those
previously employed in this industry between the
two groups is not statistically significant at
the 5 level. This means that - (a) the study was not properly done
- (b) comparability of cases and controls has been
confirmed - (c) the observed difference might be
statistically significant at the 1 level - (d) the observed difference might be readily
explained by sampling variation - (e) none of the above
62Question 7.
- (d) the observed difference might be readily
explained by sampling variation
63Question 8.
- The association between Pancreatic Carcinoma
and Serum Factor 42 (SF42) is assessed in a
case-control study with the results reported as
Risk Ratio 2.87 (Plt0.01). The best
interpretation of this statement is - (a) There is no association between pancreatic
carcinoma and SF42 - (b) SF42 causes pancreatic carcinoma
- (c) There is an association but is probably
coincidental and therefore deserves no further
consideration - (d) There is evidence of an association between
pancreatic carcinoma and SF42 - (e) It is impossible to draw any inference about
association because the data were obtained from a
case-control study
64Question 8.
- (d) There is evidence of an association
between pancreatic carcinoma and SF42
65Question 9.
- The following table demonstrates the
rates of cholera in two groups, males and
females. How many more times likely are females
at risk of getting cholera than males. - Â Yes
No Total - Female 40 60
100 - Male 20
40 60 - Note 40/60 0.67 40/100 0.4 60/100
0.6 20/40 0.50 20/60 0.33 40/60 0.67
0.4/0.33 1.21 0.67/0.6 1.12 0.50/0.33
1.51 0.50/0.4 1.25 0.67/0.33 2.03 - A) 1.21
- B) 1.12
- C) 1.51
- D) 1.25
- E) 2.03
66Question 9.
67Question 10-14.
- A cohort is conducted to evaluate the
relationship between exposure to solid foods at
an early age and the development of asthma. In
the study, 1000 infants who had solid food
introduced before 4 months of age are compared to
1,000 infants who had solid food introduced after
6 months of age. The results are shown below. - Introduction of Health Status
- Solid Food Asthma No Asthma Total
- Early 200 800
1000 - Late 100 900
1000 - Total 300 1700
2000
68Question 10.
- What is the risk of asthma in the group that had
early introduction of solid foods? - 0.10
- 0.15
- 0.20
- 0.25
- 0.35
69Question 10.
70Question 11.
- What is the risk of asthma in the group that had
late introduction of solid foods? - 0.10
- 0.15
- 0.20
- 0.25
- 0.35
71Question 11.
72Question 12.
- What is the risk ratio (early introduction of
solids versus late) for the occurrence of asthma? - 0.05
- 0.50
- 1.0
- 2.0
- 3.0
73Question 12.
74Question 13.
- The point estimate for the risk ratio (2) in the
last question indicates that the risk of asthma
associated with early introduction of solids is - Decreased
- Increased
- Not affected
- Cannot be determined
75Question 13.
76Question 14.
- The 95 confidence interval for the point
estimate is 1.4 to 4.5. The correct
interpretation of the results is - A statistically a significant association exists
between early introduction of solids and
increased risk for the development of asthma at
the level of P lt 0.05. - A statistically a significant association exists
between early introduction of solids and
decreased risk for the development of asthma at
the level of P lt 0.05. - It can be concluded with 95 confidence that
early introduction of solids is protective for
the development of asthma. - Breast feeding is an important intervention to
prevent the development of asthma. - The risk of asthma is not statistically
significantly different between early and late
introduction of solids foods at the level of P lt
0.05.
77Question 14.
- A statistically a significant association exists
between early introduction of solids and
increased risk for the development of asthma at
the level of P lt 0.05.
78Question 15-19.
- A cohort study is performed to evaluate the
relationship between inflammation as measured by
high C-reactive protein and the occurrence of
myocardial infarction among women. In the study,
500 subjects with high C-reactive protein and 500
subjects with normal C-reactive protein are
studied over a 20-year period. - During the study, 50 of the women with high
C-reactive protein and 15 of the women with
normal C-reactive protein develop a newly
diagnosed myocardial infarction.
79Question 15.
- The incidence (per 10,000 person years) for
myocardial infarction among women with a high
C-reactive protein is - 15
- 25
- 30
- 50
- 60
80Question 15
81Question 16.
- The incidence rate (per 10,000 person years) for
a myocardial infarction for a person with normal
C-reactive protein is - 15
- 25
- 30
- 50
- 60
-
82Question 16.
83Question 17.
- The incidence rate ratio for myocardial
infarction is - 0.9
- 1.0
- 2.3
- 3.3
- 5.0
84Question 17.
85Question 18.
- The risk difference is
- 0.005
- 0.007
- 0.07
- 0.05
- 1.0
86Question 18.
87Question 19.
- The attributable risk percent is
- 25
- 35.5
- 50
- 70
- 90
88Question 19.
89Study Designs
90Study Design
- Rank strongest to weakest study design
- Experimental study (strongest)
- Prospective cohort study
- Historical cohort study
- Case-control study
- Cross-sectional study
- Case-series
- Case report
91Descriptive Studies
- Correlational Studies
- Case Reports Case Series
- Cross-sectional Surveys
92Descriptive Studies
- Describe patterns of disease in relation to
variables such as person, place and time - Allow efficient allocation of health resources
- Less expensive than analytic studies because they
use information already collected e.g. vital
statistics, health survey data etc. - Used to formulate research questions
- Can not use this study to test hypotheses !
93Correlational/ Ecological Studies
- First step in investigating a possible
exposure-disease relationship - Can be done quickly and inexpensively using
existing data such as - vital statistics, hospital discharge data,
product consumption
data, environmental data
94(No Transcript)
95Cigarette Smoking and geographical variation in
CHD mortality in the US (J Chronic Dis.
20769,1967)
96The Relation of Alcohol to CHD and
Mortality Implication for Public Health Policy.
J Pub Health Policy 1198,1980
97National Mortality Rates for Coronary Heart
Disease and Malignant Growths of the Intestine,
Excluding Rectum (WHO List 1969)
98The Relation of Alcohol to CHD and Mortality
Implication for Public health Policy. J Pub
Health Policy 1198,1980
99Analytic study showing that the relationship
between alcohol and CHD is not linear as
suggested by the correlational study
100(No Transcript)
101Limitations of Correlational Studies
- Inability to link exposure with disease
- Inability to control for confounders
- Correlational data represent average exposure
levels, not individual values - Outliers can have a strong effect on the
correlation coefficient - May not reveal complex associations
- Can not prove an etiologic association !
- Ecologic fallacy
102Case Reports
- Describe the experience of a single patient or
group of patients with the same diagnosis - The commonest type of study published
- Document unusual medical occurrences
- May lead to the identification of a new disease
103Examples of Case Reports...
- 1974 3 cases of angiosarcoma of the liver among
workers at a vinyl chloride plant - 1980 5 cases of Pneumocystic carinii pneumonia
(PCP) in young gay men - 1981 case series of Toxic Shock Syndrome in
young women - 1981 Multiple cases of Kaposis Sarcoma in gay
men
104Toxic Shock Syndrome in The United States
105Problems with Case Reports
- They can not test for the presence of a valid
statistical association - Based on the experience of only one or a few
patients - The presence of a risk factor may only be
coincidental - There is no comparison group
106Case Series of Oral Contraceptive Use and
Hepatocellular Carcinoma (Age 26-35)
No.
107- Case series shows that 4 X as many women with
hepatocellular carcinoma are oral contraceptive
users - Because there is no comparison group it is
impossible to determine if the rate of OCP use is
any different from that in the healthy population - No conclusion can be made from this study
regarding the association between OCP and
hepatocellular carcinoma
108Cross-Sectional Surveys
- Also called Prevalence Surveys
- Exposure and disease are measured simultaneously
- Provides a snapshot of the population
- Examples include NHANES HHANES
109Benefits of Cross-Sectional Surveys
- Used to provide information on prevalence of
disease and health outcomes - Allows administrators to assess health status and
needs of the population - Used to formulate hypotheses
110 Problems with Cross-Sectional Surveys
- Surveys gather prevalent not incident data
- Can not determine whether exposure preceded or
resulted from disease chicken or egg dilemma - Results will reflect determinants of survival as
well as etiology - Usually can not be used to test a hypothesis
111(No Transcript)
112Cross-sectional survey of coronary heart disease
among white farm owners age 40-74 by occupational
physical activity.
113The data show an association between inactivity
and CHD...Is this because inactivity leads to
heart disease or because CHD leads to inactivity?
114Analytic Studies
- Descriptive data provide the first clues in the
investigation of a cause-effect relationship,
i.e. hypothesis generation - Once the hypothesis is formulated the next step
is to test it - The analytic studies allow further analysis,
testing and ultimately the rejection or
acceptance of the hypothesis
115Types of Analytic Studies
- Observational
- Case-control
- Retrospective cohort
- Prospective cohort
- Experimental
- Randomized Control Trial
- Meta-Analysis
116Case-control Studies
117Case-control Study
- Also called a retrospective study
- People diagnosed as having a disease (cases) are
compared with persons who do not have the disease
(controls) - The purpose is to determine if the two groups
differ by exposure
118Case-control Study
PAST
PRESENT
Look for past exposure to the risk factor in
cases and controls
Select cases and controls
119Why Use the Case-control Study Design?
- Chronic diseases e.g. cancer and CVD have long
latency periods. A cohort investigation into
chronic disease may take too long - Rare diseases (low population incidence) can not
be analyzed easily using cohort approach
120Odds Ratio Relative Risk
Case Control
a
b
Yes No
Risk factor
c
d
RR a/(ab) c/(cd)
If a disease is rare (low prevalence), then a and
c will approach zero, leading to an approximation
of OR
ad bc
121Analysis of Case-control Studies
Case Control
a
b
Yes No
OR ad bc
Risk factor
c
d
122Case-control Study and the Odds Ratio
- Incidence can not be derived in a case-control
study - The odds ratio (estimate of relative risk) can
only be calculated if - The disease has a low incidence (5 or less)
- The control group is representative of the
general population with respect to frequency of
the exposure
123Hepatoma (usual prevalence 1.5 per million)
Hepatitis B 500,000
1 case found (2x usual prevalence)
No hepatitis 500,000
0 cases (expected prevalence)
30 years follow-up
Cohort study approach Cost of study 20
million Results of study not significant
124Hepatoma (usual prevalence 1.5 per million)
Case Control
40
30
Yes
Hepatitis B
20
10
No
Case-control study approach Cost of study
50,000 Results of study significant OR 2.67
125The Benefits of Case-Control Design
- For rare diseases the case-control approach is
the most cost efficient - It is more likely to produce significant results
when studying diseases of low prevalence - Allows for the evaluation of a wide range of
potential etiologic exposures - Results available in a timely fashion
126The Problems of Case-Control Design
- Disease and exposure have already occurred at the
time of the study - creates a dilemma with
temporal association - Strong potential for bias
- Confounding
- Difficulties choosing appropriate controls
127Matching
- If cases and controls are matched by usual
confounders e.g. age, sex, SES, smoking, alcohol
etc then these factors will be equal in both
groups and will not confound the association
between the variables - But, it can be very difficult and expensive to
find a perfect match for each case
128Analysis of a Matched Case-control Study
Control Exposed Unexposed
Case Exposed Unexposed
b
a
d
c
OR b/c
129Matched-pair case-control study of exogenous
estrogens and endometrial cancer NEJM.
2931164,1975.
Control Exposed Unexposed
Case Exposed Unexposed
113
39
150
15
OR b / c 113 / 15 7.5
130Cohort Studies
131Cohort Studies
- Also called
- Longitudinal studies
- Follow-up studies
- Incidence Studies
- Prospective Studies
132Cohort Studies
- The second major type of observational study
- Individuals are followed on the basis of presence
or absence of risk factors - Individuals are then followed over time to
determine if they develop a specific outcome /
disease
133Cohort Studies
100 with disease
Cohort free of disease 2000 participants
1,900 without disease
Now ---------------------------------------------
Future
134Prospective Cohort Study
- Initiation of study occurs before occurrence of
disease - Groups of exposed and unexposed individuals are
monitored over time to assess the development of
disease - The incidence of disease in both groups is
compared - Potential confounders documented
135Prospective Cohort
Select cohort classify as to exposure to factor
Follow to see if disease develops
Present
Future
136- One hundred children known to have been
exposed to high levels of lead during the first
12 months of life were followed for 15 years 40
developed an affective disorder. A similar group
of 100 children not exposed to lead were followed
over the same time period. 5 of these children
developed an affective disorder. - What is the incidence of affective disorders
among those exposed to high lead levels? - What is the relative risk for those exposed
compared to those with no exposure?
137- Affective Disorder
- present
absent - exposed
- not exposed
- What is the incidence of affective disorders
among those exposed to high lead levels? - 40/100 40 or 0.40
- What is the relative risk for those exposed
compared to those with no exposure? - 40/100 ? 5/100 8
40
60
5
95
138Advantages of Cohort Studies
- The temporal sequence between exposure and
disease is clearly established - Well suited for assessing the effects of rare
exposures - Incidence of disease can be calculated
- Can examine multiple effects of a single exposure
- True estimate of risk can be calculated
- Natural history of disease can be studied
139Disadvantages of Cohort Studies
- Usually involve large numbers of individuals over
many years, therefore expensive (average 10,000
per participant) - Subject to attrition (loss of follow up)
- Subject to outcome bias - knowledge of exposure
can influence ascertainment of disease outcome
140Nested Case-control Studies
- A case-control study can be inserted into a
cohort study - When enough individuals have developed the
outcome of interest, they can be compared to
controls - This allows interim evaluation of the association
between the variables
141Intervention Studies
142Intervention Studies
- Known as the clinical trial
- The gold standard in clinical research
- Similar to a cohort study in that individuals are
enrolled on the basis of exposure - Unlike cohort studies the exposure is allocated
by the investigator - The main difference between observational and
intervention studies is that individuals have no
control over the exposure they receive
143Design...intervention
- Parallel
- patients are randomized to each of the
intervention and non-intervention arms for the
duration of the experiment - Cross-over
- patients spend an equivalent amount time in the
intervention and non-intervention arms - Factorial
- suitable for studying the effects of more than
one intervention - especially suited for possibly synergistic
interventions
144Design...Parallel Intervention
145Design...Cross-over
146Design...Factorial
147Types of Intervention Studies
- Therapeutic
- Determine the ability of an agent or procedure to
diminish symptoms, prevent recurrence or decrease
death from disease. The disease is already
present e.g. most drug trials - Simvastatin Study
- Preventive
- Evaluation of whether the agent or procedure
decreases the development of disease in those
free from disease e.g. most vaccine trials and
behavior modification - MR FIT Study
148Issues With Intervention Studies
- Ethics
- The active assignment to a treatment or procedure
means that the study must be ethical i.e. can not
assign to treatments known to be harmful or
withhold beneficial treatment - Feasibility
- It may not be possible to do certain studies if
- the procedure is widespread (MVI and nurses
- in NHS)
- Cost
- 15 - 25 K per participant
149Conducting a Clinical Trial
- Formulate the hypothesis
- Select participants random assignment, choose
sample size - The power is the ability of a statistical test to
detect differences among comparison groups - The greater the sample size, the greater the power
150Efficacy vs. Effectiveness
- Efficacy is the ability of a treatment to work in
the trial or study setting (in a
volunteer/compliant study population with active
follow-up) - Effectiveness is the ability of the treatment to
work under realistic circumstances. Takes into
account patient compliance, acceptability of
treatment and patient diversity
151Allocation of Study Regimens
- Assignment to treatment groups should occur after
the study population is chosen and informed
consent obtained - Randomization tables and computer generated
randomization used most frequently - Block randomization used when you wish to
maintain equal numbers of e.g. women and men in
each group - Triple blinded vs. double blinded vs. single
blinded trials
152Block Randomization
Study population n1200 (100 women 1100 men)
100 women
1100 men
Randomization occurs after assignment into
blocks male and female
Drug A
Drug A
Drug B
Drug B
n50 n50
n550 n550
153Reasons for Randomization
- Reduces bias - no selection, recall or
interviewer bias should occur (double blinded) - Reduces confounding (known confounders) - should
have equal numbers of potential confounders in
all groups if the sample size is big enough - Randomization also reduces effect of unknown
confounders
154Analysis of Data
- Similar statistics used for cohort analysis
- Intention to treat (preserve random allocation,
simulates real world experience)- determines
effectiveness - Explanatory only analyse those who actually
take treatment - determines efficacy, but
vulnerable to bias
155Intention to Treat
Study population
Drug A 1000
Placebo 1000
500 compliant
500 non-compliant
1000 compliant
250 cured
250 cured
Intention to treat means that the cure rate for
Drug A is calculated as 250/1000 25 The cure
rate for placebo arm is 250/1000 25
Drug A has no effect
156Explanatory
Study population
Drug A 1000
Placebo 1000
1000 compliant
500 compliant
500 noncompliant
250 cured
250 cured
Explanatory analysis means that the cure rate for
Drug A is calculated as 250/500 50 The cure
rate for placebo arm is 250/1000 25
Drug A is twice as effective as
placebo
157- Important
- Analyses in the medical literature (drug trials)
should always be analyzed and reported by
intention to treat method - Sometimes the explanatory method is used, however
the investigators should also show the results
analyzed by intention to treat
158(No Transcript)
159Phases of Clinical Trials
- Phase 1Initial testing in humans following
animal studies. Identify dose limiting
toxicities, tolerated doses, describe
pharmacology (metabolism, excretion) - Phase 11 testing in subjects with disease to
determine activity and therapeutic efficacy.
Validate toxicity and dosage data - Phase 111 Randomized trials for comparison
160Post-licensing Evaluation
- Safety and efficacy of drug
- Disease surveillance
- Adverse effects
161Experimental Studies
- Advantages
- intervention possible
- graded intervention if desired
- randomization
- truest test of causation
- minimization of bias
- high internal validity
- possible to examine multiple outcomes
- Disadvantages
- resource intensive
- time and cost
- compliance
- ethical issues
- possibility of type 1 and 2 errors remain
- conflicting results between RCT
- generalizability
162Systematic Reviews Meta-Analysis
163What is a Systematic Review?
- Systematic Review
- Based on protocol
- Involves critical appraisal
- Synthesis of the data
- Meta-analysis
- Statistical combination of data
164Systematic vs. Narrative Reviews
165Why do a Systematic Review?
- To examine the quality of the evidence
- To summarize the treatment benefit in a given
therapeutic area provide the best estimate of
its direction and magnitude - To resolve discordance between trials
- To investigate reasons for the discordance
166(No Transcript)
167Meta-analysis
- Statistical combination of the results of
multiple studies, weighed by the inverse of the
variance of the estimate in each study (larger
sample sizes receive more weight) - Dichotomous outcomes RR, OR
- Continuous outcomes Weighted mean difference,
standardized mean difference
168(No Transcript)
169Question 20.
- Radiologists have claimed that the survival
rates after cancer surgery are improved by
pre-operative radiation of the cancer. We plan a
randomized clinical trial to perform the
following comparison in operable patients. One
group of patients will receive surgery
immediately. The other group will receive a
course of intensive radiotherapy for a month,
followed by a month of recovery. At the end of
the second month, surgery will be performed in
those patients who are still operable. The
results will be compared in those patients who
received surgery alone versus who received
radiotherapy plus surgery - (a) the comparison between patients who received
surgery alone and those who received radiotherapy
plus surgery cannot be a comparable study because
it is not possible to carry out a double blind
study - (b) the study achieved the comparison of like
with like because random allocation was employed - (c) the comparison is valid because it would be
unethical to perform surgery on in-operable cases
following the pre-operative radiotherapy - (d) for valid comparison, the radiotherapy plus
surgery group should include all the patients
allocated to this regime, including those who
became in-operable following the pre-operative
radiotherapy - (e) none of the above
170Question 20.
- (d) for valid comparison, the radiotherapy
plus surgery group should include all the
patients allocated to this regime, including
those who became in-operable following the
pre-operative radiotherapy
171Question 21.
- Epidemiologic studies of the role of a
suspected causal factor of a disease may be
observational or experimental. The essential
difference between experimental and observational
studies is that in experimental investigations - (a) the study and control groups are equal in
size - (b) the study is a cohort study
- (c) the characteristics of interest are
quantitative variables measured objectively - (d) the investigator determines who shall be
exposed to the suspected factor and who shall not - (e) controls are used
172Question 21.
- (d) the investigator determines who shall be
exposed to the suspected factor and who shall not
173Question 22.
- A double-blind study of a vaccine is one in
which - A) The study group receives the vaccine and the
control group receives a placebo - B) Neither observer nor subjects know the nature
of the placebo - C) Neither observer nor subjects know which
subject receives the vaccine and which receives a
placebo - D) Neither the study group nor the control group
knows the identity of the observers - E) The control group does not know the identity
of the study group
174Question 22.
- C) Neither observer nor subjects know which
subject receives the vaccine and which receives a
placebo
175Questions 23-32.
- A randomized, double-blind, placebo-controlled
clinical trial is conducted to determine whether
memantine, an NMDA antagonist, can reduce
clinical deterioration in patients with
Alzheimers disease. Among 252 patients
randomized in equal numbers to memantine or a
placebo, the groups were similar at baseline with
respect to demographic characteristics and level
of dementia. The study was 28 weeks, with 29
experimental patients and 42 controls not
completing the study. Based upon cognitive,
functional, and behavioral assessment, 29 of the
patients treated with memantine and 10 of
controls demonstrated a predefined favorable
clinical response. Serious adverse events
occurred among 13 of patients on memantine and
18 of controls.
176Question 23.
- The similarity of treatment groups with respect
to baseline characteristic most likely occurred
because of - Use of intention-to-treat analysis
- The placebo effect
- Use of randomization
- Use of blinding
- Use of informed consent
177Question 23.
178Question 24.
- One control in ten had a clinical response. This
is best explained by - Use of intention-to-treat analysis
- The placebo effect
- Use of randomization
- Use of blinding
- Use of informed consent
179Question 24.
180Question 25.
- Neither the patients no clinicians who evaluated
them knew individual treatment assignments. This
may be described as - Use of intention-to-treat analysis
- The placebo effect
- Use of randomization
- Use of blinding
- Use of informed consent
181Question 25.
182Question 26.
- The risk of not responding clinically among the
memantine patients was - 0.10
- 0.29
- 0.71
- 0.79
- 0.90
183Question 26.
184Question 27.
- The risk of not responding clinically among
controls was - 0.10
- 0.29
- 0.71
- 0.79
- 0.90
185Question 27.
186Question 28.
- The risk ratio of not responding clinically among
the patients treated with memantine compared to
the corresponding reference risk among controls
was - 0.10
- 0.29
- 0.71
- 0.79
- 0.90
187Question 28.
188Question 29.
- The risk ratio of a serious adverse effect among
patients treated with memantine compared to the
corresponding reference risk among control was - 0.13
- 0.18
- 0.72
- 0.82
- 0.87
189Question 29.
190Question 30.
- The results suggest that compared to with
controls, patients treated with memantine have - Better clinical response and fewer serious
adverse effects - Better clinical response but more serious adverse
effects - Worse clinical response but fewer serious adverse
effects - Worse clinical response and more serious adverse
effects - The same clinical response but few serious
adverse effects
191Question 30.
- A. Better clinical response and fewer serious
adverse effects
192Question 31.
- Although more controls than patients treated with
memantine failed to complete the study, all
patients were analyzed according to the original
assignment. This best described as -
- Use of intention-to-treat analysis
- The placebo effect
- Use of randomization
- Use of blinding
- Use of informed consent
193Question 31.
- A. Use of intention-to-treat analysis
194Question 32.
- Some patients may have enrolled with cognitive
deficits great enough to impair their
understanding of the purpose and risk of the
study. In these instances, caregivers represented
the patients in accepting the risks and benefits
of the study. This process is best described as - Use of intention-to-treat analysis
- The placebo effect
- Use of randomization
- Use of blinding
- Use of informed consent
195Question 32.
- E. Use of informed consent
196Question 33.
- A study is conducted in a country to evaluate the
prevalence of type 2 diabetes - Case-control study
- Cohort study
- Clinical trial
- Cross-sectional survey
- Meta-analysis
- Ecologic study
197Question 33.
- D. Cross-sectional survey
198Question 34.
- Patients with newly developed breast cancer are
asked about previous dietary intake or fat. Their
responses are compared to patients admitted to
hospital for plastic surgery procedures. - Case-control study
- Cohort study
- Clinical trial
- Cross-sectional survey
- Meta-analysis
- Ecologic study
199Question 34.
200Question 35.
- Two hundred patients with rheumatoid arthritis
are randomly assigned to 6 months of a new
anti-inflammatory drug or standard care - Case-control study
- Cohort study
- Clinical trial
- Cross-sectional survey
- Meta-analysis
- Ecologic study
201Question 35.
202Question 36.
- A study evaluates the per-capita intake of
calcium and the prevalence of hypertension in 8
different countries. - Case-control study
- Cohort study
- Clinical trial
- Cross-sectional survey
- Meta-analysis
- Ecologic study
203Question 36.
204Question 37.
- The results of several case-control studies of
the association between exposure to rug shampoo
and the development of Kawasaki disease are
examined to produce a summary conclusion. - Case-control study
- Cohort study
- Clinical trial
- Cross-sectional survey
- Meta-analysis
- Ecologic study
205Question 37.
206Question 38.
- In 1945, 1,000 women were identified who
worked in a factory painting radium dials on
watches. The incidence of bone cancer in these
women up to 1975 was compared to that of 1,000
women who worked as telephone operators in 1945.
Twenty of the radium dial painters and four of
the telephone operators developed bone cancer
between 1945 and 1975. This study is an example
of a - (a) cohort study
- (b) experimental study
- (c) clinical trial
- (d) cross-sectional study
- (e) case-control study
207Question 38.
208Question 39.
- A hypothetical study of 500 patients
hospitalized with a pathological confirmed
diagnosis of a Ca breast. For each case a
control patient without cancer of the breast was
selected and matched by race and five year age
group. Both cases and controls were provided
with a questionnaire which included a question on
history of x-ray exposure. This is an example
of - (a) an experimental study
- (b) a cohort study
- (c) a case control study
- (d) a clinical trial
- (e) a cross sectional study
209Question 39.