Epidemiology E2200b

Dr. John Koval

Professor of Biostatistics

Department of Epidemiology & Biostatistics

University of Western Ontario

With thanks to

Dr. Mark Speechley

Professor of Epidemiology

Department of Epidemiology & Biostatistics

Course Objectives

You will be able to:• understand methodological foundations of applied

human health research• critically appraise original articles about things that are

claimed to be ‘good’ and ‘bad’ for us• perform fundamental calculations using published data• discuss why studies of the same question can get

different answers (and why this doesn’t mean the science is flawed)

• list the bases for criticisms and misunderstandings of the science of epidemiology (know the true rather than the imagined limitations)

Objectives of Lectures 1&2

You will learn definitions, key concepts, history, modern applications

You will be able to:• correctly use some terminology• describe historical roots, evolution of modern

epidemiology• recognize epidemiology as a basic science

for clinical medicine, public health, health services research, outcomes research, etc.

Epidemiology: Informal definition

The branch of medical science that helps us identify factors that:

– Keep us healthy (part of ‘health promotion’)– Make us sick (etiologic research)– Help us get better again (therapeutic

research)

“Identifying factors” is NOT the same as “understanding causal mechanisms”

Some Results

Epidemiological methods have discovered numerous

causal factors of health outcomesThese findings underlie:

massive behavioural change after 1950evidence-based health care & health policy

Disagreements among studies are inevitable and do not signify weaknesses of the methodology.

Identify factors that:

• Keep us healthy: physical activity, fruits and vegetables in diet, vitamins and minerals, clean air and water, vaccines

• Make us sick: deficiencies of the above factors; smoking; (some) bacteria, viruses, parasites

• Help us get better again: pharmaceuticals, surgery, rehabilitation

Putative (potential) causal factors precede causal mechanisms

• Often begins with a clinical observation

1774 Dr. Percival Pott – noticed cancer of the scrotum in chimney sweeps, implicated ‘something in soot’

• Mechanistic knowledge takes years to develop;

we now know that soot contains polycyclic aromatic hydrocarbons that lead to squamous-cell carcinoma

Chimney Sweeps (con'd)

Chimney Sweeps Act :

Sweeps must be at least 8 years old

Sweeps must be provided suitable clothes– and accommodation–

Other causal risk factors that began with clinical hunches

Exposure• Cigarette smoke

Disease/outcome• Lung cancer (1940s)

Significance? Lung cancer was once very rare. Beginning of epidemic observed among soldiers who had started smoking in WWI. Became the leading cancer death.

rubella

Exposure

Maternal rubella

(red measles)

Viruses not previously known to cause birth defects (‘teratogenic’).

All women planning pregnancy now immunized

Disease/outcome

Birth defects (1940s)

What is ‘causation’?Experiment 1:Fred is exposed to A ….. [time passes] …… Fred gets Disease B [turn back the clock, hold everything else constant]

Experiment 2:Fred is not exposed to A … [same time passes] … Fred does not get Disease B

We can define a causal exposure as i) one that is followed by a disease outcome ii) that would not have occurred had the exposure not occurred iii) all else held constant.

√ The perfect research design. √ Proves 100% causal certainty in individuals. (Unfortunately, we cannot reverse time.)

Dr. Mark’s Magic Potion(A late night infomercial)

• Hi, Friend. Want to ace your grades in university? Well, Dr. Mark has been teaching for years and has concocted a Magic Study Potion in his kitchen laboratory. If it doesn’t increase your marks by one full letter grade, return the unused portion of the product and I’ll cheerfully refund the unspent portion of your money!….And that’s not all!!....Order now and you’ll receive absolutely free….

Evaluating causal claims

• The Magic Potion claims to causally increase students’ grades.

.

• Is there a way to prove with 100% certainty that any student’s grade was or was not affected by the Magic Potion?

Causal Certainty Necessity and Sufficiency Criteria

Necessary Cause: The Magic Potion is necessary for increased grades: only students who took my potion increased their marks by a full letter grade; none others did.

Sufficient Cause: The Magic Potion is sufficient for increased grades: every student who took my potion increased their grades.

A perfect correlation!

Took potion

Increased one letter grade Yes No

Yes Sufficiency(all exposed have outcome)

n/a

No n/a Necessity (no unexposed have outcome)

How many biological, psychological or sociological causes can you name that meet both necessity and sufficiency criteria?

How many can you name that meet even one of these criteria?

Causation and Correlation

Causation occurs when factor A leads to (or causes) factor B

Correlation happens when factor A and factor B are related, so that when factor A is present, factor B often is present,

and visa versa

“You can’t prove causation with correlation”

• True, but:– You don't need to know the exact cause before doing

something– We don’t need to understand a causal mechanism to

act to reduce exposure– Dozens of examples exist where epidemiologic

associations have subsequently been demonstrated to be causal.

– If an association is causal, every day we fail to act out of scientific prejudice, people will needlessly get ill or even die.

We will be wrong sometimes.

Example: in an investigation of an outbreak of Hepatitis A

hot dog sausages were implicated.

The whole consignment of sausages was thrown out.

People were spared from the disease, although the actual mechanism was not clear.

It turns out that the source of the bacteria was actually the relish.

“You can’t prove causation with correlation” is true, but…

• You can’t prove causation without correlation either.

• All identified causes began with observed correlations.

The problem isn’t correlation, it’s failure to control for CONFOUNDING – other explanations that could account for the correlation.

How to prove causation?ApproachesBest: expose Person A,

observe; go back in time, remove exposure, observe and compare

2nd Best: random assignment to exposure (Experimental)

3rd Best: observe people in different exposure groups

(Observational)

LimitationsCan’t do time travel:“counterfactual”

Unethical with negative outcomes

Often impractical (time)

Potential for confounding*

*Latin, confundere (pour together; confuse)

2nd best:The RCT (Random Controlled Trial)• Randomize students to Magic or Placebo Potion:

All known and unknown factors are distributed by chance

• Collect data on factors that could affect grades, compare two groups at baseline, should be similar as the sample size increases

• If imbalanced, can statistically adjust final estimates• Observe between-group difference in grades

3rd best: Observational Designs• Are not true experiments• People select themselves into exposures • Unknown or unmeasured factors

(confounders) could be the true cause of any observed difference

• As our theory improves (as we can explain a larger portion of the variation in outcomes) so does our ability to estimate the true causal effect of any single factor

The role of confounding

Cigarette Smoking

DiseaseCoffee consumption

Non-causalassociation:heavy smokerstend to be heavy coffeedrinkers

True causal effect

Spurious association

Smoking, a true cause of disease, will confound (bias) the association between coffee and disease. The apparent association

with coffee is due to the correlation between coffee and smoking.

Confounding (con'd)If you measure association of smoking and cancer

in the presence of a measurement of coffee consumption, the true effect of smoking will be diminished

Coffee consumption is a confounder of the Smoking – lung cancer relationship

Determination of actual risk factor and actual confounder depends on other (clinical) studies

Malaria (‘bad air’): A classic case of confounded association

Swamps (musty air) Malaria

Highlands (fresh air)No

Malaria

Confounder____________

True cause

Spurious association

Solution; leave swanp, What’s the true cause (vector) of malaria?

Epidemiology* (definition)

• “the study of the occurrence and distribution of health-related states or events in specified populations, including the study of determinants influencing such states, and the application of this study to control health problems” (Porta M. A Dictionary of Epidemiology, 5th ed, 2008:81). (emphases added)

*From Greek; epi (upon) dēmos (people), logos (word, reason)

“distribution” (Porta, 2008)“The complete summary of the frequencies of the

values or categories of a measurement made on a group of persons. The distribution tells either how many or what proportion of the group was found to have each value (or each range of values) out of all the possible values that the quantitative measure can have”.

Usually presented broken down by characteristics such as person, place, and time.

Age distribution of percentage of pregnancies ending in miscarriage/stillbirth, by age of women at

end of pregnancy, Canada, 1974 and 1992

0

5

10

15

20

25

Allages

15-19 20-24 25-29 30-34 35-39 40-44

1974

1992

Source: Health Reports, Summer 1996, 8:13

%

“determinants”

“any factor that brings about change in a health condition or other defined characteristic. (Porta, 2008).

Identifying possible (and probable) causal factors is not the same as explaining causal mechanisms

If a factor is causal, reducing exposure will reduce outcome even if we don’t understand the mechanism

“study” can be:

• Surveillance – (e.g. mandatory disease reporting)

• Descriptive (hypothesis generating)– (e.g. proportion of pregnancies that end in

miscarriage/stillbirth, by characteristics of person, place and time)

• Analytic (hypothesis testing)– (estimates of X-Y association from

observational studies)

• Experiments (clinical trials)

Analytic Epidemiology: Primary role is etiologic*

Exposures ('determinants')

For example,– Physical (ionizing

radiation)– Chemical (lead)– Biological (needlesticks)

– Social: educational attainment, poverty

– Behaviours: tobacco, diet

Outcomes (‘health related states and events’)

For example,• Diseases with biological

models

• Illnesses without biological models

• Injuries• Birth outcomes• Psychological states such as

QOL (Quality of Life)*Greek, aitia (cause)

Key concept: Reliable case definition

• Case definition: A set of criteria (not necessarily diagnostic criteria) that must be fulfilled in order to identify a person as a case of a particular disease (Porta, 2008:32)

– Clinical or Laboratory criteria or both– Scoring systems with points that match disease

features (e.g. Multiple sclerosis)

• Reliability: The degree to which the results obtained by a measurement or procedure can be replicated (Porta, 2008:214)

Key concept: Risk

RISK(def): The probability that an event will occur, e.g., that an individual will become ill or die within a stated period of time or age. (Porta, 2008:217)

Major aim of Epidemiology is to quantify the risk of developing disease or other negative health state posed by various exposures (molecules, microorganisms, environments, behaviors).

Probability

• Causation of health and illness is extremely complex

• Even widely agreed upon causes fail to meet necessity and sufficiency:– “Grandma smoked a pack a day and died peacefully

in her sleep at 110, and Uncle Elmo got lung cancer and never smoked”.

• We need to rely on probability statements: the probability of an outcome is 2, 3, 4.. times higher among exposed than unexposed

Observed versus predicted probability

Average (predicted) risks estimated from groups, used to advise individual patients: (e.g. risk of adverse surgical outcome; risk of cancer recurrence)

But! individuals will either have (risk = 100%) or not have (risk = 0%) an outcome over a specified time period (you can’t have ‘32% of a stoke’).

– Estelle, 28, never-smoker, former Varsity volleyball player, has a stroke. Observed individual risk of stroke for that year = 100%

– Jerome, 75, high blood pressure, smoker, does not have a stroke. His observed individual risk for that year = 0%

People like Estelle face a very low predicted risk; people like Jerome face a much higher predicted risk

History of Epidemiology

Epidemiology is a young science with ancient roots in the study of epidemics (def: “The occurrence in a community or region of cases of an illness, specific health related behavior, or other health-related events, clearly in excess of

normal expectancy.” Porta, 2008:79) “Clearly in excess” differs by disease and time

frame (e.g. H1N1 or Lung Cancer)

Began with communicable diseases; methods have been adapted for chronic diseases and other health states and events (injuries, birth outcomes, etc)

Demons, Miasms* and Germs

Epidemiologic insights (e.g. events are not random) are clear in the writings of Hippocrates 2500 years ago.

Millenia passed before we had the intellectual foundation to scientifically test 2 competing hypotheses about the causes of epidemic diseases

Key period: 1850s England: Drs. John Snow (cholera) and William Budd (typhoid fever)

*From Greek, miainein (to pollute).

2 theories of epidemic disease

Miasmatic (miasma)• Air has a ‘bad quality’• Rotting organic matter• ‘Miasma’ could be

passed from cases to susceptibles in contagious diseases

Contagion• Invisible entities• Spread through direct

contact, droplet spread or contaminated fomites

Most physicians supported miasma; it explained the facts better:• didn’t know that asymptomatic people could be infectious

(‘well carriers’)• Didn’t know about immunity

1850s England: Urbanization, industrialization, poverty, crowding,

filth and epidemic disease

Increasingly scientific medical profession continued to favour miasmatic theory over contagion:

• London, 1854: Cholera epidemicwas it miasma or germs?

• How to prevent/stop epidemic

Lack of Sanitation

“Miasms”

GermsDisease

William Farr Deaths from Cholera in 10,000 Inhabitants by Elevation of

Residence above Sea Level, London, 1848-1849Elevation above Sea Level (ft) Number of Deaths

<20 120

20-40 65

40-60 34

60-80 27

80-100 22

100-120 17

340-360 8

Data from Farr W: Vital Statistics: A Memorial Volume of Selections from the Reports and Writings of William Farr (edited for the Sanitary Institute of Great Britain by Noel A. Humphreys). London, The Sanitary Institute, 1885.

John Snow, M.D. (1813-1858)

www.ph.ucla.edu/epi/snow.html

• 1847- theory that cholera is communicable and waterborne

• Used spot maps of cases’ residences, compared to location of public water pumps

> 500 cholera fatalities within 250 yards of Cambridge and Broad Streets in a 10 day period.

Snow's diagram

Snow (con'd)

Eventually convinced Parish authorities to remove Broad Street pump handle during August-September 1854 epidemic

Modern day view

The pump on

Broadwick street

Plus the Pump Pub

Cases of Cholera by date of onset, London, Aug. 19 – Sept. 19, 1854

0

20

40

60

80

100

120

140

160

Fatal attacksDeaths

Epidemic curve adapted from Roht et al, 1982:300

Pump handleremoved

August September

f

“Natural Experiment”London England, ~1853

• 2 major water suppliers: Lambeth, and Southwark & Vauxhall

• Lambeth moved their intake to a cleaner section up river

• Interviewed household members to ascertain which of two companies supplied their water

• Compared 1853 cholera cases according to water company (retrospective study)

Cholera mortality by water supply, 1st seven weeks of epidemic

(Roht et al, 1982:304)

Water Supply

# houses Cholera Deaths

Deaths/10,000 houses(risk)

Southwark & Vauxhall

40,046 1,263 315.4

Lambeth Co.

26,107 98 37.5

Rest of London

256,423 1,522 59.4

Epidemiologic measures of association: Relative Risk*

One form of Relative Risk = Risk Ratio

Deaths/10,000 exposed (S&V) = 315.4 = 8.4Deaths/10,000 unexposed (Lambeth) 37.5

Mortality was 8.4 times more common in S&V houses than in Lambeth houses.

Based on these non-experimental (non-randomized) findings, who here would choose S&V?

The Establishment reacts:

• The Lancet: “not by any means conclusive”• Royal College of Physicians: “theory as a whole is

untenable” … continued to support “foul or damp air” as the cause

• Board of Health Medical Inspectors: “We see no reason to adopt this belief” (1854)

• “far-fetched doctrine”(Chapman, 1866)• 1884, Robert Koch (Nobel, 1905) identified Vibrio

Cholerae, made no mention of Snow’s work

This is, unfortunately, not uniqueMany epidemiologic findings, even after multiple

replications and systematic testing and rejection of bias explanations, are stubbornly resisted. Why?

• economic self-interest • resistance to behavioral change • unwillingness to admit past practices killed people

Unfortunately, isolated first findings are often given the most sensationalistic media coverage