Basic Concepts In Epidemiology
Amit Khera, MD, MSc June 7, 2006
Epidemiology
the study of the distribution and determinants of disease frequency
MacMahon, B: Epidemiology: Principles and Methods, 1970.
John Snow (1813-1858)
Father of epidemiology London cholera epidemic of 1854 Broad Street Pump
Epidemiology
EXPOSURES
determinants of disease (association)
OUTCOMES
endpoints of a study (disease)
Outcome Measures
Prevalence proportion- proportion of a population with the outcome (disease) at a single point in time
Incidence- the number or proportion of individuals developing the outcome (disease) during a period of time
incidence proportion (risk) incidence rate
Obesity Among U.S. Adults 2002
No Data
<10%
10%14%
15%19%
20%24%
25%
Source: Behavioral Risk Factor Surveillance System, CDC
Lipid Levels and Heart Disease
14-y incidence rates (%) for CHD 14 12 10 8 6 4 2 0
< 40 4049 5059
60
260 230259 200229 < 200
HDL-C (mg/dL)
Castelli WP et al. JAMA 1986;256:28352838
Linking Exposures to Outcomes
Outcome measures are descriptive characteristics about distribution of the outcome
ex: what is the prevalence of lung cancer?
Measures of Association
quantifying relationships
Measures of Association
Difference Measures Risk Difference (absolute risk reduction)=
Incidence exposed - Incidence unexposed
NEJM 2004;350:1495-1504
4162 subjects with acute coronary syndromes randomized to standard dose v. high dose statin therapy followed for mean of 24 months outcome- incidence of death, MI, revascularization, unstable angina, or stroke Incidence of outcome in exposure group: 22.4% Incidence of outcome in control group: 26.3% --------risk difference= -3.9%
Measures of Association
Ratio Measures Risk Ratio Incidence Rate Ratio Hazard Ratio Odds Ratio
Relative Risk
Incidence exposed/Incidence unexposed
JAMA 2004;291
community randomized trial in Kenya to see if insecticide-treated bednets could reduce childhood morbidity and mortality
Children 1-11 months Incidence rate of death treatment group: 100/1000 person-years Incidence rate of death control group : 128/1000 person-years --------relative risk (RR) of death= 0.78 in treated group Relative Risk Reduction= 1-RR 22%
Are There Alternative Explanations?
Measures of Validity Internal Validity Chance- (p-value) Bias Confounding
External Validity Generalizability
Bias
Bias- systematic error affecting the results of the study
Selection bias- association between disease and exposure occurs because of the way participants were selected, not by underlying truth Recall bias- occurrence of outcome results in increased recall of exposures
ex: maternal recall bias
Informational bias- differential misclassification of exposure or outcome (MD Behavior Bias)
Selection Bias
What is the prevalence of depression in patients with congestive heart failure? CHF (exposure) Depression measured by questionnaire (outcome)
STUDY A
Patients in a CHF clinic were approached to be involved in the study 52% were found to have depression
STUDY B
Patients were randomly selected from a populationbased study of CHF 23% were found to have depression
Confounding
Confounding- mixing of the effect of an exposure on the outcome with the effect of another exposure Ex: Downs Syndrome
1.8
Affected Babies/1000
9 8 7 6 5 4 3 2 1 0
1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 1 2 3 4 Birth Order 5
<20
20-24
25-29
30-34
35-39
40+
Maternal Age
External Validity
Generalizability- how well do these results apply to other populations? Ex: Framingham Heart Study
Ten-Year Prediction of CHD Events in CMCS Men and Women Using the Original Framingham Functions
Liu, J. et al. JAMA 2004;291:2591-2599.
Study Types
Observational
cohort (follow-up) case-control cross-sectional (prevalence)
Experimental
randomized trial
Cohort Study
Cohort study- study that follows or traces any designated group over a period of time Classify subjects by exposure
Follow for outcome
Benefits: -less bias -can estimate population rates of disease or exposure specific risk Drawbacks: -requires large population, especially for rare outcome -can require long follow-up period
JAMA 2004;291:2448-2456.
High-risk patients at an urban county hospital enrolled 190 cocaine exposed infants and 186 non-exposed infants outcome Wechsler Preschool and Primary Scales of intelligence at 4 years
190 cocaine-exposed infants 190 non-exposed infants
4 years
outcome outcome
4 years
RESULTS: no difference in full-scale verbal or performance IQ scores
Case-Control Study
Study in which subjects with the outcome (cases) are compared to those without (controls) to determine different exposure distribution (usually retrospective) Classify subjects by outcome Determine exposure
Benefits: -good for rare disease (outcomes), long latency -requires fewer subjects than cohort study Drawbacks: -can introduce bias in selection of controls -cannot estimate population rates of disease or exposure specific risk
HMG-CoA Reductase Inhibitors and the Risk of Hip Fractures in Elderly Patients
JAMA 2004;283:3211-3216
Reviewed histories from patients enrolled in New Jersey
Medicare or Medicaid or Pharmacy Assistance for Aged and Disabled Program 1222 patients who had a hip fracture 4888 control patients selected without hip fracture (4:1matched for age and sex)
1222 with hip fracture 4888 without fracture exposure (statins) exposure (statins)
RESULTS: statin use: 2.2% cases v. 4.4% controls Odds Ratio of hip fracture with statin use- 0.50
Cross-sectional Study
Study used to assess the prevalence of disease at one point in time
JACC 2004;43:1791-1796
90% 80% 70% 60% 50% 40% 30% 20% 10% 0% <10% 10-20% >20% Framingham Risk Category
Prevalence
Randomized Controlled Trial
Type of cohort study in which the exposures are assigned Gold standard for epidemiologic trials Randomization ensures equal distribution of confounders
1. Randomization
2. Assign Exposure
Gender= known confounder =unknown confounder
Low-dose statin
Subjects with ACS High-dose statin
24 months
Outcome
26.3%
24 months
Outcome
22.4%
Randomized (Exposure Assigned)
Suggested Reading
Rothman K.J. Epidemiology: An Introduction, New York: Oxford University Press, 2002. Hennekens CH, Buring JE. Epidemiology in Medicine, Boston: Little, Brown and Co., 1987
