Epidemiology. Bis vivit qui bene vivit

Transcription

1 Epidemiology Bis vivit qui bene vivit

2 What is Epidemiology? Epidemiology, literally translated from Greek, means "the study of [a] people". Epidemiology is the study of how a disease is distributed in a population and of the factors that influence or determine this distribution.

3 Objectives of Epidemiology - determine the extent of disease in a population - study the natural history and prognosis of a disease - characterise the aetiology of a disease - identify risk factors or protective factors - evaluate preventive and therapeutic measures - provide the foundation for developing public policy and regulatory decisions

4 Historical Examples Edward Jenner ( ) - smallpox survivors were immune to the disease - variolation, the administration of material from infected people, was common (dangerous) practice - cowpox (mild form of smallpox) was found in milkmaids, who never contracted smallpox - pus from cowpox-infected patient was used to perform the first successful smallpox vaccination - miasmatic theory of cholera: cloud of disease close to earth, with lower altitudes more susceptible than higher ones - house to house survey of where cholera deaths obtained their water - proof of contagious nature and transmission pathways John Snow ( )

5 Point Prevalence Morbidity Measures Prevalence (π) proportion of affected individuals present in a population at a specific time point mathematically: probability π t that an individual randomly drawn at a specific time point is affected Period Prevalence proportion of affected individuals present in a population during a specific time period mathematically: probability π d that a randomly drawn individual is, or has been, affected during a specific time period

6 Period and Point Prevalence time 30 years period prevalence 4 ˆd π = = point prevalence 3 ˆt π = =

7 Confidence Interval The number X of diseased individuals in a sample of size n follows a Bin(n,π) distribution. KI: πˆ ± t1 α/2,n 1 π ˆ (1 πˆ) n yields a confidence interval for the estimate of π. Example: 20 diseased among 35 probands πˆd = = 95%KI: ± 2.04 = 0.57 ±

8 Morbidity Measures Incidence Proportion (γ), "Risk" A: number of new cases in a population at risk that occur during a specified time period N: number of individuals at risk during a specified time period γ = A N mathematically: probability (or risk) that an unaffected, randomly chosen individual gets affected during the time period of interest

9 Incidence Proportion time T: 30 years 3 ˆ γ = =

10 Morbidity Measures Incidence Rate (γ), "Risk" N: number of individuals at risk during time period A: number of new cases arising during time period T i : time units spent under risk by the i th individual γ = A N i = 1 Ti mathematically: (time) rate at which unaffected, randomly chosen individuals get affected

11 Incidence Rate time T: 30 years 3 γˆ = = incidents per person-year

12 Prevalence and Incidence D: disease duration 1-π γ prevalence pool π 1 E(D) expected prevalence pool inflow γ ( 1 π) t expected prevalence pool outflow 1 π t E(D)

13 Prevalence and Incidence In a stable, closed population (i.e. without migration into, or out of, the population) γ ( 1 π) t = 1 π t E(D) so that 1 π π = γ E(D) Through causing longer disease duration, improved medical care may increase the disease burden to society in the form of an increased prevalence.

14 Typhus Pediculus humanus Rickettsia prowazekii In the German population, the incidence rate of typhus is per year. The average disease duration is approximately one month. 1 π π = γ E(D) = = In the German population, approximately =14 cases of typhus are to be expected at any point in time.

15 Prevalence and Incidence Problems - ambiguous or incorrect diagnoses, latency - identification of highly selected cases from hospital admissions (severity, policy) - bad recording of cases (incomplete, missing) - variable diagnostic standards (temporal, regional) - ambiguous definition of population base (medical, ethnic, social) - temporal changes of disease patterns (spatial, phenotypical)

16 Effect Measures Relative Risk (ρ) Let a population be stratified into two strata (e.g. "exposed", "not exposed") with corresponding incidence rates or proportions ("risks") γ e and γ n during the observational period. ρ = γ γ e n is called the 'relative risk' under exposure. ρ>1: "risk factor", ρ<1: "protective"

17 Relative Risk (ρ) exposed not exposed time 30 years 5/ ˆ ρ = = = 2/

18 Clinical Trials Types of Epidemiological Studies Experimental (Interventional) - performed on single diseased individuals in a clinical setting - evaluation of therapeutic measures (e.g. drugs) Field Trials Assignment of Exposure by Investigator - performed on single non-diseased individuals in the field - evaluation of preventive measures (e.g. vaccination) Community Interventions - performed on groups of non-diseased individuals - evaluation of preventive measures (e.g. water treatment)

19 not exposed exposed Archetypal Experimental Study time

20 Cohort Studies Types of Epidemiological Studies - performed prospectively on non-diseased individuals of known exposure status, disease incidences are recorded Case-Control Studies Non-Experimental (Observational) Assignment of Exposure by Nature - performed retrospectively on individuals of known disease status, exposure status is recorded Cross-Sectional (Prevalence) Studies - performed retrospectively on the whole population or on a representative sample, disease and exposure is recorded

21 Archetypal Observational Study time

22 The Framingham Study Objective To identify the common factors or characteristics that contribute to cardiovascular disease (CVD) by following its development over a long period of time in a large group of participants who had not yet developed overt symptoms of CVD or suffered a heart attack or stroke. Design In 1948, 5209 men and women between the ages of 30 and 62 were recruited from Framingham, Massachusetts (representing 2/3 of the adult population). In 1971, another sample of 5135 men and women was established, comprising the offspring of the original cohort and their spouses.

23 The Framingham Study Results Careful monitoring of the Framingham Study population has led to the identification of the major CVD risk factors - high blood pressure - high blood cholesterol - smoking - obesity - diabetes - physical inactivity and critical information upon related factors such as age, gender, and psychosocial issues. The Framingham study has produced approximately 3500 articles in leading medical journals.

24 Risk and Odds Odds = 1 Risk Risk "The risk of catching a viral flue this winter is 0.20." one affected among five at risk "The odds of catching a viral flue this winter is 0.25." one affected for every four non-affected

25 Horse Race Betting "Old Mule" payoff odds 1:5 1:10 1:50 1:200 fair poor poor poor good fair poor poor good good fair poor good good good fair

26 Effect Measures Odds Ratio (OR) OR = γ γ e n /(1 /(1 γ γ e n ) ) If risks γ e and γ n are "sufficiently small" for the chosen time unit, i.e. of the order a few percent, then OR = γ γ e n /(1 /(1 γ γ e n ) ) γ γ e n = ρ

27 Odds Ratio (OR) exposed not exposed time 10 years time unit: 10 years 2 / 8 OR = = 1/ /10 ρˆ = = 1/

28 Odds Ratio (OR) not exposed exposed time 30 years time unit: 30 years 5/5 5/10 OR = = 4.00 ρˆ = = /8 2/10

29 Which Effect Measure? Cohort Studies (Relative Risk) affected not affected total exposed a b a+b not exposed c d total a+c b+d c+d n a a A e n and + b Ne c + d Nn a/(a c/(c + + b) d) = ˆγ ˆγ e n c = ρˆ A

30 Which Effect Measure? Case-Control Studies (Odds Ratio) affected not affected total exposed a b a+b not exposed c d total a+c b+d c+d n a c A A e n and b d N N e n A A e n a/c b/d (N e A /An A )/(N e A )... ˆγ ˆγ n /(1 ˆγ /(1 ˆγ e e e = = = n n n ) ) OR

31 Effect Measures Confidence Intervals ρ ˆ = a/(a c/(c + + b) d) OR = a/c b/d σˆ ln( ρ ) = 1 a a 1 + b + 1 c c 1 + d 1 a 1 b 1 c σˆ ln( OR) = d confidence intervals for the natural logarithms ln( ˆ) ρ ± z 1 σˆ α/2 ln( ρ) ln( OR) ± α/2 z 1 σˆ ln(or)

32 Cohort Study affected not affected total exposed not exposed total a/(a + b) 10/150 a/c 10/5 ρˆ = = = 2.00 OR = = = c/(c + d) 5/150 b/d 140/145 95% CI: % CI:

33 Case-Control Study affected not affected total exposed not exposed total a/c 100/50 OR = = = b/d 140/ % CI: a/(a + b) 100/240 ˆ ρ = = = c/(c + d) 50/

34 Which Effect Measure? Case-control studies do not normally allow the estimation of relative risks. An odds ratio provides a good approximation of the relative risk for a disease if the incidence rate (over the chosen time unit) is small.

35 Attributable Risk Aetiological Fraction Q: Which incidences are due to the exposure? A: This question cannot be answered on the basis of epidemiological data alone. E.g. many smokers will get lung cancer from causes other than tobacco smoke (e.g. asbestos, radiation, chance).

36 Attributable Risk Rate Fraction Q: What proportion of the risk under exposure is due to the exposure? AR = γ e γ γ e n = ρ ρ 1 assesses the excess risk in an individual.

37 Attributable Risk (AR) γ e,males = 0.50 γ n,males = 0.20 γ e,females = 0.08 γ n,females = 0.02 ρ = 0.50/ ρ = 0.08/ males = females = AR males = = 0.60 AR females = = Although the disease risk of exposed males is much higher than the disease risk of exposed females, the AR is higher in females because their relative risk is higher.

38 Population Attributable Risk Excess Fraction Q: What proportion of the incidences in the population is attributable to the exposure? PAR = γ γ γ n = f e fe ( ρ 1) ( ρ 1) + 1 γ: general incidence, f e : exposure frequency assesses the excess morbidity in a population.

39 Population Attributable Risk (PAR) γ e,males = 0.50 γ n,males = 0.20 γ e,females = 0.08 γ n,females = 0.02 ρ = 0.50/ ρ = 0.08/ males = females = f e,males = 0.20 f e,females = PAR males = = 0.23 PAR females = = Although the AR is higher in females than in males, the PAR's are the same because males are exposed more frequently than females.

40 Summary - Epidemiology is the science that studies the distribution of diseases and of their causal factors in populations. -The major morbidity measures used in epidemiology are the prevalence, i.e. the disease frequency, and the incidence, i.e. the occurrence rate of new cases. -Epidemiological studies can be either interventional or observational. In terms of their timing, studies can be of prospective or retrospective design. -The effect of an exposure upon disease risk can be measured by the relative risk or the odds ratio. -(Observational, retrospective) case-control studies do not allow the estimation of relative risks, only of odds ratios.

41 Appendix Standardisation Let a population be stratified into k strata (e.g. age, gender) with incidence rates γ 1,...,γ k. Let s 1,...,s k be "standard" person-times, e.g. in a reference population. γ S = k s γ i= 1 i i k s i= 1 i is called the 'standardised incidence rate'.

42 Appendix: Sex-Specific Incidence Rates population females males a Σt i ˆγ f a Σt i ˆγ m y y y y ˆ1 γ = 10/300 = ˆ2 γ = 11/600 = y -1 y -1

43 Appendix: Sex-Specific Incidence Rates population females males a Σt i ˆγ f a Σt i ˆγ m y y y y y -1 ˆ1 γ = 10/300 = ˆ2 γ = 11/600 = y -1 s female =150 s male =100 ˆγ 1, S = ( ) /250 = ˆγ 2, S = ( ) /250 = y -1 y -1

44 Appendix: Collapsing of Relative Risks γ e,males = 0.50 γ n,males = 0.20 γ e,females = 0.08 γ n,females = relative risk overall females 2.5 males proportion of males

45 Appendix: Collapsing of Odds Ratios γ e,males = 0.50 γ n,males = 0.20 γ e,females = 0.08 γ n,females = females odds ratio overall males proportion of males