Cohort, case-control & cross- sectional studies Kostas Danis EPIET Introductory course Menorca, Spain 16/9-12/10/2012 Source: Alain Moren, EPIET Introductory courses

Epidemiological studies Two types Observation Experiment

Unethical to perform experiments on people Exposed Disease occurrence Exposure assigned Not exposed Unethical to perform experiments on people if exposure is harmful

If exposure not harmful Treatment Preventive measure (vaccination) Randomised Controlled Trial Blinded Doses Time period Risk - effect No bias If RCT not possible Left with observation of experiments designed by Nature Cohort studies Cross-sectional studies Case control studies

Cohort studies marching towards outcomes

What is a cohort? One of 10 divisions of a Roman legion Group of individuals sharing same experience followed up for specified period of time Examples EPIET cohort 2012 birth cohort cohort of guests at barbecue occupational cohort of chemical plant workers influenza vaccinated in 2011-12

follow-up period

Calculate measure of frequency Cumulative incidence incidence proportion attack rate (outbreak) Incidence rate end of follow-up

Cohort studies Purpose Cohort membership Study if an exposure is associated with outcome(s) Estimate risk of outcome in exposed and unexposed cohorts Compare risk of outcome in two cohorts Cohort membership Being at risk of outcome(s) studied Being alive and Being free of outcome at start of follow-up

Cohort studies unexposed exposed

Incidence among unexposed Cohort studies exposed Incidence among exposed Incidence among unexposed unexposed

Presentation of cohort data: 2x2 table ill not ill Total ate ham a b a+b did not eat ham c d c+d Risk in exposed= a/a+b Risk in unexposed= c/c+d

Number of NEW cases of disease Total person - time of observation Incidence rate Number of NEW cases of disease Total person - time of observation It is more precise because it includes the time at risk in the denominator. The incidence rate is the number of new cases divided by the total number of person time observed

Number of NEW cases of disease Total person - time of observation Incidence rate Number of NEW cases of disease Total person - time of observation Rate Denominator: - is a measure of time - the sum of each individual’s time at risk and free from disease It is improtant to know, when using the term rate - a measurement of time is included in the denominator and That only persons are included in the denominator until they either develop disease or they are not at risk for developing disease any more (i.e diseased, moved out, immune, lost, …) So the denominator is the sum of each individuals time at risk and free from disease Lets say what it means on a graph

Person-time x x A B C D E 6.0 10.0 8.5 5.0 Total years at risk 35.5 90 91 92 93 94 95 96 97 98 99 Time at risk A B C D E 6.0 10.0 8.5 5.0 x x Assume a population of 5 persons Person A is followed from 1990 until 1996 … because he may die or be lost to follow up… he will contribute 6 years to our observation Person B is born in 1992 and develops disease in 1997.. He or she will contribute 6 years to our observation Person C contributes to the entire observation period of 10 years Person D moves to the study area in 1991 and leaves the area in mid 1998 .. He will contribute for 8,5 yrs Person E develops disease after 5 yrs So our denominator will be the sum of the time of each individualat risk when they are free of disease which in our example are 35.5 years Total years at risk 35.5 -- time followed x disease onset

Incidence rate (IR) (Incidence density) 90 91 92 93 94 95 96 97 98 99 00 Time at risk IR = 2 / 35.5 person years = 0.056 cases / person year = 5.6 cases / 100 person years = 56 cases / 1000 person years A B C D E 6.0 10.0 8.5 5.0 x x This total person time observed will become our denominator and the number of persons who diseased are in the numerator Therefore the incidence rate or incidence density will be 2/35,5 person years which equals 0.056 cases per person year Which is the same as 5.6 cases per 100 person years Or expressed as 56 cases per 1000 person years Note that the incidence density is not limited and may reach infinite values Total years at risk 35.5 -- time followed x disease onset

Presentation of cohort data: Person-years at risk Tobacco smoking and lung cancer, England & Wales, 1951 Person-years Cases Smoke 102,600 133 Do not smoke 42,800 3 Source: Doll & Hill

Presentation of data: Various exposure levels

Prospective cohort study Disease occurrence Exposure Study starts time time Exposure Study starts Disease occurrence

Retrospective cohort study Disease occurrence Study starts Exposure time

Recipe: Cohort study Identify group of Measure incidence of disease exposed subjects unexposed subjects Measure incidence of disease Compare incidence between exposed and unexposed group

Incidence among unexposed Cohort studies exposed Incidence among exposed Incidence among unexposed unexposed

Effect measures in cohort studies Absolute measures Risk difference (RD) Ie - Iue Relative measures Relative risk (RR) Rate ratio Risk ratio Ie Iue Ie = incidence in exposed Iue= incidence in unexposed

Cohort study Non cases Risk % Total Cases Exposed 100 50 50 50 % 50 50 50 % Not exposed 100 10 90 10 % Risk ratio 50% / 10% = 5

ill not ill Incidence ate ham 49 49 98 50 % did not eat ham 4 6 10 40 % Risk difference 50% - 40% = 10% Relative risk 50% / 40% = 1.25

Interpretation of Risk Ratios RR>1 RR=1 RR<1 Risk factor No association Protective factor 26

Does HIV infection increase risk of developing TB among drug users?

Vaccine efficacy (VE) VE = 1 - RR = 1 - 0.28 = 72% So, a vaccine efficacy of 72% means that 72% of the cases which would have occurred among the vaccinated would they have been unvaccinated. have been prevented. VE = 1 - RR = 1 - 0.28 = 72%

Various exposure levels Population Cases Incidence at risk a1 High N1 I1 a2 Medium N2 I2 a3 Low N3 I3 c Unexposed Nne Iue

Various exposure levels Population Cases Incidence RR at risk a1 High N1 I1 RR1 a2 Medium N2 I2 RR2 a3 Low N3 I3 RR3 c Unexposed Nne Iue Reference

Cohort study: Tobacco smoking and lung cancer, England & Wales, 1951 Source: Doll & Hill

Disadvantages of cohort studies Large sample size Latency period Cost Time-consuming Loss to follow-up Exposure can change Multiple exposure = difficult Ethical considerations

Strengths of cohort studies Can directly measure incidence in exposed and unexposed groups true relative risk Well suited for rare exposure Temporal relationship exposure-disease is clear Less subject to selection biases outcome not known (prospective)

Cohort studies Can examine multiple effects for a single exposure Population Outcome 1 Outcome 2 Outcome 3 exposed Ne Ie1 Ie2 Ie3 unexposed Nne Iue1 Iue2 Iue3 RR1 RR2 RR3

A cohort study allows to calculate indicators which have a clear, precise meaning. The results are immediately understandable.

Cross-sectional (prevalence) studies

Cross-sectional studies Observation of a cross-section of a population at a single point in time Recruitment of study participants Population Population sample Observation for the presence of: One or more outcomes One or more exposures

Sampling Sampling Population Target Population Sample When we want to take a sample we first need to define our target population. The target population is the population about which you want information, that you wish to make conclusions about from the results of the study. The study or sampling population is the population from which the sample (sampling frame) is drawn (the population from which you select your sample). It may be a more limited, an accessible population. For example, suppose you want to estimate the prevalence of flu-like symptoms in a country and you conduct telephone interviews; Your target population will be the total population in the country and the study population will be all people with telephones. Or if you want to estimate the vaccination coverage among 6 year old children in Spain and you take a sample of school children; your target population is all 6 year olds in Spain, whereas your sampling population is all first Grammar class school children. Target Population

Uses of cross-sectional surveys in public health Estimate prevalence of disease or their risk factors Estimate burden Measure health status in a defined population Plan health care services delivery Set priorities for disease control Generate hypotheses Examine evolving trends Before / after surveys Iterative cross-sectional surveys

Potential objectives of a cross sectional study Descriptive Estimate prevalence Analytic Compare the prevalence of a disease in various subgroups, exposed and unexposed Compare the prevalence of an exposure in various subgroups, affected and unaffected

Presentation of the data of an analytical cross sectional study in a 2 x 2 table Ill Non ill Total Exposed a b a+b Non exposed c d c+d Simultaneous measurement of outcomes and exposures

Cross-sectional study Non cases Prevalence % Total Cases Exposed 1,000 500 500 50 % Not exposed 1,000 100 900 10 % Prevalence ratio (PR) 50% / 10% = 5

Measuring association in analytical cross-sectional surveys Prevalence among exposed / prevalence among unexposed Prevalence ratio Formula equivalent to risk ratio Concept different No incidence Only prevalence depends on both occurrence of new cases & duration of disease

Prevalence of West Nile virus (WNV) infection by place of residence, Central Macedonia,Greece, 2010 Infected Total Prevalence ratio Rural 38 491 7.7% 5.9 Urban 3 232 1.3% Ref

Prevalence of HIV infection by socioeconomic status, African country X, 1999 Infected Total Prevalence ratio High class 15 235 6.4% 2.6 Low class 11 450 2.4% Ref Women of higher socioeconomic status are more likely to survive longer with HIV

Prevalence of hepatitis C (HCV) infection by quantity of therapeutic injections, Hazabad, Pakistan, 1993 No.of injections Infected Total Prevalence ratio >10 9 41 22% 22 0-10 4 52 8% 8 1 82 1% Ref Another potential explanation for the association between injections and HCV infection is that HCV infected individuals became symptomatic from their infection and so sought more health care, and so received more injections.

Advantages of cross-sectional surveys Fairly quick Easy to perform Less expensive Adapted to chronic diseases

Limitations of cross-sectional surveys Limited capacity to document causality (exposure and outcome measured at the same time -difficult to establish time sequence of events) Not useful to study disease etiology Not suitable for the study of rare / short diseases Not adapted to severe / acute diseases Not adapted to incidence measurement.

Limitations of causal inference in analytical cross sectional studies Prevalent cases Exposure and outcome examined at the same time

Principle of case control studies

Our objective is to compare the incidence rate in the exposed population to the rate that would have been observed in the same population, at the same time if it had not been exposed

Source population Exposed Unexposed

Source population Cases Exposed Unexposed

Source population Cases Exposed Sample Unexposed Controls

Cases Controls Source population Exposed Sample Unexposed Controls: Sample of the denominator Representative with regard to exposure Controls

Intuitively if the frequency of exposure is higher among cases than controls then the incidence rate will probably be higher among exposed than non-exposed

Case control study Exposure ? Disease Controls Retrospective nature

Distribution of cases and controls according to exposure in a case control study Cases Controls Exposed a b Not exposed c d Total a + c b + d % exposed

Distribution of cases and controls according to exposure in a case control study Cases Controls Exposed a b Not exposed c d Total a + c b + d % exposed a/(a+c) b/(b+d)

Distribution of myocardial infarction by oral contraceptive use in cases and controls Oral Myocardial contraceptives Infarction Controls Yes 693 320 No 307 680 Total 1000 1000 % exposed 69.3% 32%

Distribution of myocardial infarction by amount of physical activity in cases and controls Physical Myocardial activity Infarction Controls >= 2500 Kcal 190 230 < 2500 Kcal 176 136 Total 366 366 % exposed 51.9% 62.8%

Volvo factory, Sweden, 3000 employees, Cohort study 200 cases of gastroenteritis Water Cases Controls Consumption YES 150 ? NO 50 ? Total 200 200

Two types of case control studies Exploratory New disease New risk factors Several exposures "Fishing expedition" Analytical Define a single hypothesis Dose response

Cohort studies Case control studies Rate/risk Rate/risk difference Rate Ratio/Risk ratio (strength of association) Case control studies No calculation of rates/risks Proportion of exposure Any way of estimating measures of association?

Odds Probability that an event will happen Probability that an event will not happen Probability that cases/controls will be exposed Probability that cases/controls will not be exposed The odds are often used in horse races and are a ratio The odds are the probability that an event will happen divided the probability that an event will not happen

a/c b/d Case control study Cases Controls Odds ratio a b Exposed OR= (a/c) / (b/d) = ad / bc Not exposed c d a+c Total b+d % exposed a/(a+c) b/(b+d) d/(b+d) % unexposed c/(a+c) Odds of exposure a/c b/d

Case control study Cases Controls Odds ratio 50 20 4 a b Exposed 50 20 4 a b Exposed Not exposed 50 80 c d Total 100 100 OR= (a/c) / (b/d) = ad / bc = (50x80) / (20x50) = 4 Odds of exposure 50/50 20/80

Case control study design Cases Controls Odds ratio E a b a b a x d ---- --- = --- ---- c d b x c c d E

Cases Controls Odds ratio Ate chicken 181 251 2.1 Did not eat chicken Frequency of chicken consumption in campylobacter cases and controls, Republic of Ireland and Northern Ireland, 2003 Cases Controls Odds ratio Ate chicken 181 251 2.1 Did not eat chicken 15 44 Ref

Cases Controls Odds ratio Contact with dog Yes 29 93 0.40 No 158 201 Frequency of contact with a dog in campylobacter cases and controls, Republic of Ireland and Northern Ireland, 2003 Cases Controls Odds ratio Contact with dog Yes 29 93 0.40 No 158 201 Ref

Distribution of myocardial infarction by recent oral contraceptive use in cases and controls Oral Myocardial contraceptives Infarction Controls OR Yes 693 320 4.8 No 307 680 Ref. Total 1000 1000 Odds 693/307= 320/680= of exposure 2.2 0.5

Distribution of myocardial infarction by amount of physical activity in cases and controls Physical Myocardial activity Infarction Controls OR >= 2500 Kcal 190 230 0.64 < 2500 Kcal 176 136 Ref. Total 366 366 odds of 190/176= 230/136= exposure 1.1 1.7

Distribution of cases of endometrial cancer by oestrogen use in cases and controls Oestrogen use Cases Controls Odds ratio High a1 b1 a1d/b1c Low a2 b2 a2d/b2c None c d Reference

Relation of hepatocellular adenoma to duration of oral contraceptive use in 79 cases and 220 controls Months of OC use Cases Controls Odds ratio 0-12 7 121 Ref. 13-36 11 49 3.9 37-60 20 23 15.0 61-84 21 18.1 >= 85 49.7 Total 79 220 Source: Rooks et al. 1979

Advantages of case control studies Rare diseases Several exposures Long latency Rapidity Low cost Small sample size Available data No ethical problem

Limitations of case-control studies Cannot compute directly risk Not suitable for rare exposure Temporal relationship exposure-disease difficult to establish Biases +++ control selection recall biases when collecting data Loss of precision due to sampling

The cohort study is the gold-standard of analytical epidemiology CASE-CONTROL STUDIES HAVE THEIR PLACE IN EPIDEMIOLOGY, but if cohort study possible, do not settle for second best

Thank you!

Back-up slides

} } a c E a P1 I1 = a / P1 c P0 E I0 = c /P0 I1 = -------- P1/10 E a Population denominator Cases E a P1 } I1 = a / P1 a/P1 I1/ I0 = ------ c/P0 c P0 E I0 = c /P0 Population sample Cases a I1 = -------- P1/10 E } a P1 /10 a/P1 I1/ I0 = ------ c/P0 c I0 = -------- P0/10 c P0 /10 E

} Source population E a P1 I1 = a / P1 c P0 E I0 = c /P0 E a b c d E Cases Pop. E a P1 I1 = a / P1 } a/P1 I1/ I0 = ------ c/P0 c P0 E I0 = c /P0 = sample Cases Controls E a b P1 b --- = ---- P0 d c d E

} Source population E a P1 I1 = a / P1 c P0 E I0 = c /P0 E a b c d E Cases Pop. E a P1 I1 = a / P1 } c P0 E I0 = c /P0 a/P1 a . P0 a . d I1 / I0 = ------ = ------- = ----- = c/P0 c . P1 c . b a / c ------ b / d = sample Cases Controls E Since d/b = P0 / P1 a b P1 b --- = ---- P0 d c d E