introduction to epidemiology -...
TRANSCRIPT
1
Introduction to epidemiology
PhD course Spring 2010University of Copenhagen
Anders Koch, afdelingslæge Ph.D. MPHStatens Serum Institut
The first and today’s lecture: Introduction to epidemiology
• Background, definition and change over time
• Statistical associations and sources of errors
• Causal relationships
Fathers of epidemiology
John Snow1813 – 1858
Cholera in London 1849-
Peter Ludwig Panum1820 – 1885
Measles in the Faroe Islands 1846
2
Cholera epidemics England 1831-1854
Lancet, "History of...the...cholera in England and Scotland". 1831-32
King Cholera dispenses contagion: The London cholera epidemic of 1854
The medical profession…
"Long life to our Central Board . . . May we preserve our health by bleeding the country . . .”
George Cruikshank (1792-1878): The Central Board of Health: Cholera Consultation (London: S. Knight, 1832)
John Snow 1813 - 1858
• Obstetrician in Frith Street, London
• Considered cholera to becaused by polluted water
• Prevailing theory breathingin of vapour or contagioussubstance in the athmosphere (miasma)
3
Water supply in London 1854
Southwark & Vauxhall (green)
Lambeth(red)
Cholera in London 1854
• This argumentation without effect on authorities or water works
• Loss of definitive proof
591,422256,423Rest of London
379826,107Lambeth Company
3151,26340,046Southwark and Vauxhall
Company
Deaths/10,000 houses
Deaths from cholera
No. of houses
Broad Street, Soho 1854
4
The pump in Broad Street
• Cholera outbreak 19. aug. – 30. sept. 1854
• 616 dead
• Sick persons short distance to particular pump
• Most sick persons direct access to pump
• Snows microscopy: White, fluffy particles in water
• Widow in Hampstead who had died from cholerahad daily her waiter get water from Soho pump
Intervention
• Anecdote: Snow sneaked out at night and removed pump handle making the epidemic stop
• Reality: Handle removed by public health authorities onSnows Snows suggestion September 8th; the removalhad no effect on epidemic
Snows epidemiology
• ’… it is obvious that no experiment could have been devisedwhich would more thoroughly test the effect of water supply onthe progress of cholera than this…. To turn this experiment to account, all that was required was to learn the supply of water to each individual house where a fatal attack of cholera mightoccur.’
• Theory about spread of infectious diseases in general and specifically about the spread of cholera before knowledge of the cause of cholera
• Concepts– Randomisation (rich/poor, males/females, children/elderly)– Mortality rates– Intervention
5
Snow’s tracks in history
Society for Epidemiological Research, slogan competition 1981:
Epidemiologists Love Snow Jobs
Objectives of epidemiology course
Become acquainted with epidemiological terminology
Promote
• Understanding & interpretation of epidemiologic data
• Good epidemiologic research
• Understanding of decisions made on epidemiologic data
Netdoktor, februar 2001
6
Definition
• Epi (on) demos (people, population) logos (knowledgeof) = The knowledge of what happens to people
• ’The study of the distribution and determinants of healthrelated states or events in specified populations, and the application of this study to control of healthproblems’ (Last 1988)
• ’The study of the distribution and determinantsof disease frequency ’
Epidemiology: Definition & objectives
”The outbreak in epidemic form of a disease of pseudo-scientific meticulosis. The symptoms of the condition are characterised by:
a) evidence of a certain degree of cerebral exaltation; b) an inherent contempt for thosewho cannot understand logarithms, and c) the replacement of humanistic and clinical valuesby mathematical formulae.
The systemic effects of this disease areapparent; patients are degraded from human being to pricks in a column, dots in a field, ortadpoles in a pool; with the eventualelimination of the responsibility of the doctor to get the individual back to health.”
Epidemiology is (also)…
Logic and common sense!
7
On epidemiologic studies
It is more important to increase the quality of data in the collection phase than to applysophisticated statistics
A. Bradford Hill
In general
Garbage in, garbage out!
Causes of death 1900-1982, DK
100.0%Total100.0%Total
18.0%Other7.2%Other
1.7%Diabetes1.9%Diphtheria
1.9%Chronic liver disease4.2%Accidents
2.0%Pneumonia./influenza4.5%Cancer
2.1%Suicide5.9%Nephritis
2.9%Chronic lung disease6.3%Diarrhea/enteritis
6.5%Stroke7.6%Stroke
6.6%Accidents9.4%Heart disease
23.9%Cancer11.2%Tuberculosis
34.4%Heart disease11.8%Pneumonia/influenza
19821900
8
Evolvement of epidemiology
• Study of the distribution and determinants of health-related states orevents in specified populations (Last: A Dictionary of Epidemiology)
• In the 1950’erne shift from infectious diseases to chronic diseases(coronary disease, cancer, etc.) – advent of antibiotics
• Framingham study 1949- (coronary diseases)
• Denmark in the lead (Cancer Registry, CPR)
US Surgeon General William Stewart
Talk to the Congress 1969:
The war against pestilence is over and now it is time to close the book on infectiousdisease
Result:
Low priority to infectiousdisease and microbiologicalresearch in Western Europeand the USA
Epidemiology in a historical perspective
• Patterns of mortality and morbidity are changing– From ”infections” to ”chronic” disorders
• Diseases have different natural histories– From diseases with short latency periods (weeks to years)– To diseases with long latency periods (years to decades)
• Changing effects of determinants– From great to moderate effects
• Requires development of new methodology
9
1982
Lancet. 1982 May 15;1(8281):1083-7. Related Articles,Links
Risk factors for Kaposi's sarcoma in homosexual men.
Marmor M, Friedman-Kien AE, Laubenstein L, Byrum RD, William DC, D'onofrio S, Dubin N.
An investigation of 20 homosexual men with histologically confirmed Kaposi's sarcomaand 40 controls revealed significant associations between Kaposi's sarcoma and use of a number of drugs (amyl nitrite, ethyl chloride, cocaine, phencyclidine, methaqualone, and amphetamine), history of mononucleosis, and sexual activity in the year before onset of the disease. Patients with Kaposi's sarcoma also reported substantially higher rates of sexually transmitted infections than did controls. Multivariate analysis indicatedindependent significant associations for amyl nitriteand sexual activity and showed useof phencyclidine, methaqualone, and ethyl chloride to be non-significant. Evaluated at the median exposure for patients, the analysis yielded risk-ratio estimates of 12.3 for amyl nitrite (95% confidence limits 4.2, 35.8) and 2.0 for sexual activity (95% confidence limits 1.3, 3.1).
2009
Swine flu symptom checker:
If you wake up looking likethis, don’t go to work
Miranda Carnewro, 18, and Jorge Juarez, 18, wears a masks as they wait to clear U.S. Customs crossing from Ciudad Juarez, Mexico, into El Paso, Texas, Monday, April 27, 2009. (AP Photo/LM Otero)
Epidemiological way of thought(Infectious diseases)
• Is there a problem ?• What characterises the problem?
– When– Where– Who
• Hypothesis
• Is the hypothesis correct?
• Devise public health measures
}}
aÉëÅêáéíáîÉ
ÉéáÇÉãáçäçÖó
^å~äóíáÅ
ÉéáÇÉãáçäçÖó
10
Descriptive & analytic epidemiology
Randomised intervention
Epidemiologic focus
Case control study
Cohort study
Case report
Case series
Ecological study(correlational study)
Cross sectional study
generating testing
Hypothesis-
Incidence report
Basic assumption
DiseaseCause
Basic assumption
Lung cancerSmoking
11
What is a cause?
• What is the cause of being run down by a car on H.C. Andersens Boulevard?
• What is the cause of a post worker in New Jersey catchingAnthrax?
• What is the cause of children being admitted to hospital with RSV infektion ?
• What is the cause why Anders Koch did not catchsalmonella-infection at the New Year Dinner given by the Medical Association of Copenhagen in Domus Medicayear 2000?
Causes of infection
`ÜçäÉê~ qìÄÉêÅìäçëáë
(Robert) Koch’s postulates
• Organism present in every case of disease• Organism may be isolated and grown in pure culture
• Organism must cause disease if afflicted on a susceptiblelaboratory animal
• Organism must be isolated and identified from the laboratoryanimal
• Antrax demonstrated by these rules
However
• What if the organism cannot be cultured (bacteria are dead)?• What if the organism cannot be grown?
12
Practical epidemiology (infectiousdiseases)
• Why do you need to know the cause of a disease?
• In order to intervene!
• Therefore the necessary step is to know as much of the causal chain as to be able to intervene, but not necessarily all elements (e.g. cholera and TB)
Association and cause
• In a study it has been observed that a certainfactor characterizes the sick persons
• If this factor appears more frequent amongthe sick than expected, the factor is associated with the disease
• Is the suspected factor a cause?
Formally
• Association refers to the statistical dependencebetween two variables, that is, the degree to whichthe rate of disease in persons with a specificexposure is either higher or lower than the rate of disease among those without that exposure
• A causal association is one in which a change in the frequency or quality of an exposure or characteristicresults in a corresponding change in the frequency of the disease or outcome of interest.
13
Bradford Hills criteria for causality
1. Strength2. Consistency
3. Specificity4. Temporality
5. Biological gradient6. Plausibility
7. Coherence8. Experimental evidence
9. Analogy
Bradford Hills criteria rearranged
Is there a valid statisticalassociation?
• Is there a strong association?• Is there consistency with other
studies?• Is there biological credibility to
the hypothesis?• Is the time sequence
compatible?• Is there evidence of a dose-
response relationship?
Can this valid statistical associa-tion be judged as cause & effect?
• Is the association likely to bedue to chance?
• Is the association likely to bedue to bias?
• Is the association likely to bedue to confounding?
Observation
Coffee Pancreatitis(Exposure Outcome)
lÇÇë=ê~íáç=Ñçê=ÇêáåâáåÖ ÅçÑÑÉÉ Ñçê=é~åÅêÉ~íáíáë Å~ëÉë
NM=Åçãé~êÉÇ ïáíÜ Åçåíêçäë
14
Disease and cause
• Is coffee drinking really associated with (a cause of) pancreatic cancer?
`Ü~åÅÉ
_á~ë=EëóëíÉã~íáÅ ÉêêçêF
`çåÑçìåÇáåÖ
Coffee and pancreatic cancer
– Chance?• Random error• Statistical strength/significance
• 2 patients, 2 controls (OR 10, 95% CI 0.2 – 113)
– Bias?• Systematic error• Cancer cases, controls patients with other gastrointestinal
diseases (ulcers, etc. – don’t drink coffee because of disease)
– Confounding?• Unintentional mixture of effects of other factors• 10 times higher alcohol consumption among pancreatitis cases
P-values and confidence intervals
• P-value (probability)– Probability that a test statistic would be as extreme as or more
extreme than observed if the null hypothesis were true– One number only (’p=0.001’)– Reflects sample size and magnitude of effect– Large sample or large difference in estimate– Qualitative measure which evaluates one theory alternative to
another
• Confidence interval:– The computed interval with a given probability (e.g. 95%) that the
true value of a variable is contained within the interval– ’95% CI RR 1.6 – 3.2’– Combined impression of effect and statistical significance
15
Bias definition
”Any trend in the collection, analysis, interpretation, publication or review of data thatcan lead to conclusions that are systematicallydifferent from the truth”.
J. M. Last, 2001: A Dictionary of Epidemiology, Oxf University Press
U-land undersøger…..
Når intervieweren mærker at personen i den anden ende af røret er ved at miste interessen, må man sikre sig, at røret ikke bliver smækket på:
"Så ved man at det måske regner eller sner, hvor kunden bor, og så snakker man lidt om det, mens man skynder sig selv at udfylde hvad man regner med kunden ville have svaret."
Skulle folk alligevel smække røret på, fortsætter de mest rutinerede interviewere alligevel:
"Man lader som om kunden stadig er i røret - man taler videre, så de andre ved siden af ikke opdager det, og udfylder hvad man regner med kunden ville have sagt."
http://www.econ.ku.dk/milhoj/stik/uland%20unders%C3 %B8ger.htm
Confounding
`çÑÑÉÉ eÉ~êí=ÇáëÉ~ëÉ
pãçâáåÖj~íÅÜÉë=áå=éçÅâÉí iìåÖ Å~åÅÉê
pãçâáåÖ
• Confundere (latin): To mix together• Mixture of an effect of exposure on outcome with the
effect of a third factor• Presence of a factor which is predictor of outcome
and associated with exposure
`çåÑçìåÇÉê
bñéçëìêÉ lìíÅçãÉ
16
Excercise
Obesity
Stress
Cardiovascular disease
Inherited factors
Hypertension
Smoking
Shaving and all-cause mortalityConfounding?
The relation between frequency of shaving and all-cause and cardiovascular disease mortality, coronary heart disease, and stroke events was investigated in a cohort of 2,438 men aged 45–59 years….
……Men who shaved less frequently had fully adjusted hazard ratios (adjusted for testosterone, markers of insulin resistance, social factors, lifestyle, and baseline coronary heart disease) of 1.24 (95% confidence interval (CI): 1.03, 1.50) for all-cause mortality, 1.30 (95% CI: 0.99, 1.71) for cardiovascular disease mortality, 1.08 (95% CI: 0.61, 1.92) for lung cancer mortality, 1.16 (95% CI: 0.90, 1.48) for coronary heart disease events, and 1.68 (95% CI: 1.16, 2.44) for stroke events.
Ebrahim et al., Am J Epidemiol, 2003
17
Validity and reliability
Bradford Hills criteria
Is there a valid statisticalassociation?
• Is there a strong association?
• Is there consistency with otherstudies?
• Is there biological credibility to the hypothesis?
• Is the time sequencecompatible?
• Is there evidence of a dose-response relationship?
Can this valid statistical associa-tion be judged as cause & effect?
• Is the association likely to bedue to chance?
• Is the association likely to bedue to bias?
• Is the association likely to bedue to confounding?
18
Interpretation of epidemiological data
�Magnitude of effect
�Great effect hardly unknown confounder
• Is there consistency with other studies? • Have others made similar observations?
• Biologic credibility
• Is the time sequence sound?• Does exposure precede outcome?
• Is there evidence of a dose-response pattern?
Interpretation of epidemiological data
�Magnitude of effect
�Great effect hardly unknown confounder
• Is there consistency with other studies? • Have others made similar observations?
• Biologic credibility
• Is the time sequence sound?• Does exposure precede outcome?
• Is there evidence of a dose-response pattern?
19
Biological credibility?
• Personal characteristics and skull shape (phrenology)• Stress and gastric ulcers• Swimming one hour after eating
’In earlier times we thought that this disease wascaused by an evil spirit. Now we know better – it is caused by a garden gnome…’
Interpretation of epidemiological data
�Magnitude of effect
�Great effect hardly unknown confounder
� Is there consistency with other studies? �Have others made similar observations?
�Biologic credibility
• Is the time sequence sound?• Does exposure precede outcome?
• Is there evidence of a dose-response pattern?
Time sequence
Beaglehole et al., 1993
20
Interpretation of epidemiological data
�Magnitude of effect
�Great effect hardly unknown confounder
� Is there consistency with other studies? �Have others made similar observations?
�Biologic credibility
� Is the time sequence sound?�Does exposure precede outcome?
• Is there evidence of a dose-response pattern?
http://www.cdc.gov/tobacco/sgr/sgr_1964/1964%20SGR%2 0Chapter%209.pdf
Ylitalo et al., Lancet, 355; 2194-8
21
Summary day 1
• Epidemiology the study of the distribution and determinants of diseasefrequency
• Change from infectious diseases to chronic diseases in 20th Century (but infectious diseases are not irrelevant…)
• Knowledge of (necessary) cause(s) necessary for intervention
• Statistical association different than cause
• Chance, bias, and confounding must be evaluated to determineassociation
• Further factors determine causality (Bradford Hills criteria)
And the next time…
Descriptive and analytical epidemiologyand much, much more…