Upload
francis-owens
View
215
Download
0
Tags:
Embed Size (px)
Citation preview
Prospective Studies(cohort, longitudinal, incidence studies)
Sue Lindsay, Ph.D., MSW, MPH
Division of Epidemiology and Biostatistics
Institute for Public Health
San Diego State University
Prospective Study Design
Develop disease
Do not develop disease
Develop disease
Do not develop disease
Exposed Not Exposed
Population (does not have disease)
select select
Design Considerations
• These investigations are oriented to exposure status
• The objective is to find exposures that lead to disease.
• How well can you identify exposed individuals?
• How long should your follow-up be?
• How frequently do you need to measure exposure?
• Can you ensure that study subjects do not have the disease at the beginning of the study?
Design Considerations
• Baseline characteristics of exposed and not exposed subjects should not differ. Use comparable source populations with equivalent information.
• Both groups should be equally available for follow-up.
• The degree of surveillance for disease should be equal in both groups.
• Those assessing for disease should be blind to risk factor group.
Cohort Study Analysis
• Incidence rate
• Relative Risk with Confidence Intervals
• Risk Difference
• Attributable Risk
• Population Attributable Risk
The 2 X 2 Table for Cohort Studies
a
Disease
Develops No Disease
Exposed (+)
Not
Exposed (-)
b
c d
Totals
a + b
c + d
First Select
Incidence Rate
Incidence
Per 1,000=
No. of New Cases of Disease
In a Population
No. of Persons at Risk of
Developing Disease
During a Specified Period of Time
Relative Risk
Risk of disease among exposed
Risk of disease among the not exposed
OR
Risk (exposed) Incidence (exposed)
Risk (unexposed) Incidence (unexposed)
Relative Risk in a 2 X2 Table
a
Disease No Disease
Exposed (+)
Not
Exposed (-)
b
c d
Risk
a
c
a + b
c + d
Relative Risk = a/(a+b)
c/(c+d)
Relative Risk
• Relative risk measures the strength of an association
• The larger the relative risk, the stronger the association between
the risk factor and the disease
• Relative risk does not tell you about the actual risk of the
disease (this is measured with incidence)
Etiologic Objective
• If exposure is associated with disease, the incidence rate in the exposed group will be greater than in the not exposed group.
a
a + b>>>
c
c + d
Exposed Not exposed
Hypothetical Example:Maternal stress and preterm birth
110
Preterm birth Full term birth
Stressed
Not stressed
2,890
87 4,913
Risk
110
3,000
87
5,000
Relative
Risk=
110/3000
87/5000= 2.18
= 0.037
= 0.017
RR = 1: There is no association between the risk factor and the disease.
RR > 1: There is a positive association between the risk factor and the disease. The risk factor may be a cause of disease. Possible range 1 to infinity.
RR < 1: There is a negative association between the risk factor and the disease. The risk factor may be protective against the disease. Possible range 0-1.
Interpretation of Relative Risk
The Health Risk of Passive Smoking
• Hirayama et al. 1981
• Study Objective: To assess the health effects of passive cigarette smoking
• Japan 1965 - Data collected in 6 prefectures
• Studied the smoking habits of the spouses of:• 91,540 non-smoking wives
• 20,289 non-smoking husbands
IDENTIFY
NEW CASES OF
LUNG CANCER
Not exposed to
second
hand smoke
1965 1979
PASSIVE SMOKING IN JAPAN
14 YEAR FOLLOW-UP
(Death Certificates)
Population:
Non-smoking
spouses
Exposed to
second
hand smoke
Mortality In Non-Smoking Wives
00.2
0.40.6
0.8
11.2
1.41.6
1.8
Non-Smokers
1-14Cigs/Day
15-19Cigs/Day
>20Cigs/Day
Deaths • Shows increasing risk of
mortality with increasing number of cigarettes smoked by spouse
• Dose-response relationship
Rel
ativ
e R
isk
Mortality in Non-Smoking Husbands
0
0.5
1
1.5
2
2.5
Non-Smokers 1-14 Cigs/Day > 15 Cigs/Day
Deaths • Shows increasing risk of
mortality with increasing number of cigarettes smoked by spouse
Rel
ativ
e R
isk
Exposure to Second Hand Smoke Number Cancers Incidence15+ cigarettes/day 1,000 20 20/1,00010-15 cigarettes/day 1,000 10 10/1,000
5-10 cigarettes/day 1,000 8 8/1,000<5/day 1,000 5 5/1,000occasional (not daily) 1,000 4 4/1,000quitter 1,000 4 4/1,000never smoked 1,000 2 2/1,000TOTAL 7,000 53 7.6/1,000
Ever smoked 6,000 51 8.5/1,000
occasional,quitter,never 3,000 10 3.3/1,000
Relative Risk is Relative: Hypothetical Example
Relative Risks:
15+ cigarettes/never smoked: 20/2 = 10.00
Ever smoked/never smoked: 8.5/2 = 4.25
Daily/occasional,quitter,never: 10.8/3.3 = 3.25
The Royal College of General Practitioners (RCGP)
Oral Contraception Study
• England: Is oral contraceptive use a risk factor for cardiovascular disease?
• May 1968 - July 1969 • 23,000 oral contraceptive users identified and
recruited by physicians
• Equal number of non-users identified, matched for marital status and age
The RCGP Oral Contraceptive Study
1968 - 1969
23,000 OC
Users
23,000
Non- Users
1974
1974
1977
1977
1981
1981
Morbidity and Mortality Follow-up
RCGP Oral Contraceptive Study: Relative Risks for Cardiovascular Disease
0
1
2
3
4
5
6
Heart Disease
HypertensionIschemic
Heart Disease
Subarachnoid
Hemorrhage
Stroke
The Hepatitis B Virus Cohort Study
• Taiwan: November 1975 – June 1978
• Is Hepatitis B etiologically associated with primary hepatocellular cancer?
• 21,227 male Taiwanese government civil servants recruited
The Hepatitis B Cohort Study
Recruit 22,707
Taiwanese Men
1975- 1978
3,454
Hepatitis B +
1986
161 new cases
of primary
hepatocellular
cancer19,253
Hepatitis B -
The Hepatitis B Cohort Study
152
Cancer No
Cancer
HBsAg (+)
HBsAg (-)
3302
9 19244
Risk
152
3454
9
19,253
Relative
Risk=
152/3454
9/19,253= 98.1
Potential Bias in Cohort Studies
• Bias in Ascertainment of Outcome
• Staff responsible for the identification of disease are aware of exposure status
or hypotheses
• Diagnostic suspicion bias
Potential Bias in Cohort Studies
• Information Bias
Occurs when the extent or quality of the information obtained is different for exposed and not exposed study subjects.
Potential Bias in Cohort Studies
• Non-response and Loss to Follow-up
Occurs if non-response or loss is different for exposed compared to non-exposed study subjects.
Risk Difference
Risk (exposed)
Incidence (exposed)
Risk (unexposed)
Incidence (unexposed)
__
__
Maternal Stress and Preterm Birth:Risk Difference
110
Preterm birth Full term
Stressed
Not stressed
2,890
87 4,913
Risk
110
3,000
87
5,000
Risk difference = 0.037-0.017 = 0.020
= 0.037
= 0.017
The risk of preterm birth is increased by 0.020 for women who are stressed in their pregnancies
Attributable Risk Percent
Risk (exposed) Risk (not exposed)__
What percent of the risk in the exposed population
is due to the exposure?
Risk (exposed)
X 100
Maternal stress and preterm birth:Attributable risk percent
110
Preterm birth Full term
Stressed
Not stressed
2,890
87 4,913
Risk
110
3,000
87
5,000
Attributable risk = 0.037- 0.017 X 100 54%
= 0.037
= 0.017
54% of the total risk for preterm birth among stressed pregnant women is their stress
0.037=
Population Attributable Risk Percent(PARP, Etiologic Fraction)
Incidence (total population) Incidence (not exposed)__
The reduction in the incidence of the disease that
can be expected if we eliminate the risk factor.
X 100Incidence (total population)
Population Attributable Risk Percent(PARP, Etiologic Fraction)
This statistic is influenced by two things:
• Prevalence of the risk factor in the population
• The strength of the association between
risk factor and disease
Population Attributable Risk Percent(PARP, Etiologic Fraction)
Prevalence of the risk factor in the population
Incidenceof lungcancer
20%15%10% 5%
Smokers (25%) Non-smokers (75%)
17%
4%
What is the incidence of lung cancer in this population?
(17 * .25) + ( 4 * .75) = 7.25%
Population Attributable Risk Percent
Incidenceof lungcancer
20%15%10% 5%
Smokers (25%) Non-smokers (75%)
17%
4%
Incidence in total pop = 7.25%
Among the total population, what is the reduction in incidence that we could expect if we eliminated smoking (exposure) from the population?
7.25 - 4
PAR% = 7.25 X 100 = 44.8%
Population Attributable Risk Percent
44.8% of incidence could be reduced if smoking is eliminated
Incidenceof lungcancer
20%15%10% 5%
Smokers (75%) Non-smokers (25%)
17%
4%
What if there were far more smokers in the population?Incidence of lung cancer in this population = (17*.75) + (4*.25) = 13.75%
PAR% = 13.75 - 4 13.75
X 100 = 70.9%
Population Attributable Risk Percent
70.9% of incidence could be reduced if smoking is eliminated
Population Attributable Risk Percent(PARP, Etiologic Fraction)
The strength of the association between
the risk factor and the disease
Incidenceof lungcancer
20%15%10% 5%
Smokers (75%) Non-smokers (25%)
10%
4%
What if there were just as many smokers but they smoked less?(only 2 cigarettes per day each)
Incidence of lung cancer in this population = (10*.75) + (4*.25) = 8.5%
PAR% = 8.5 - 4 8.5
X 100 = 52.9%
Population Attributable Risk Percent
52.9% of incidence could be reduced if smoking is eliminated
Population Attributable Risk Percent: A comparison of three populations
PARP = Incidence in population - incidence in not exposed (incidence of lung cancer in not Incidence in population exposed is 4%)
Pop #1: Total population incidence = (.25*17) + (.75*4) = 7.25%
25% smokers PARP = 7.25 – 4 = 44.8%17% cancer incidence 7.25
Pop #2: Total population incidence = (.75*17) + (.25*4) = 13.75%
75% smokers PARP = 13.75 – 4 = 70.9%
17% cancer incidence 13.75
Pop #3: Total population incidence = (.75*10) + (.25*4) = 8.5%
75% smokers PARP = 8.5 – 4 = 52.9%
Only 2 a day 8.510% cancer incidence