18481

7/24/2019 18481

http://slidepdf.com/reader/full/18481 1/51

Cause and effect: theepidemiological approach

Raj Bhopal,

Bruce and John Usher Professor of Public Health,

Public Health Sciences Section,

Diision of Communit! Health Sciences,

Uniersit! of "dinburgh, "dinburgh "H#$%&

Raj'Bhopal(ed'ac'u)

mailto:[email protected]

mailto:[email protected]

7/24/2019 18481


"ducational objecties*n completion of !our studies !ou should

understand:

+he purpose of stud!ing cause and effect in epidemiolog!is to generate )noledge to preent and control disease'

+hat cause and effect understanding is difficult to achieein epidemiolog! because of the long natural histor! ofdiseases and because of ethical restraints on human

e-perimentation' Ho causal thin)ing in epidemiolog! fits in ith other

domains of )noledge, both scientific and non.scientific' +he potential contributions of arious stud! designs for

ma)ing contributions to causal )noledge'

7/24/2019 18481


Cause and effect

Cause and effect understanding is the highestform of achieement of scientific )noledge'

Causal )noledge permits rational plans and

actions to brea) the lin)s beteen the factorscausing disease, and disease itself' Causal )noledge can help predict the outcome

of an interention and help treat disease' /uote Hippocrates "To know the causes of a

disease and to understand the use of thevarious methods by which the disease may be

prevented amounts to the same thing as beingable to cure the disease".

7/24/2019 18481


"pidemiological contributions

to cause and effect

% philosoph! of health and disease' 0odels hich illustrate that philosoph!' 1rameor)s for interpreting and appl!ing the

eidence' Stud! designs to produce eidence' "idence for cause and effect in the

relationships of numerous factors and diseases'

Deelopment of the reasoning of otherdisciplines including philosoph! andmicrobiolog!, in reaching judgement'

7/24/2019 18481


% cause2

+he first and difficult 3uestion is, hat is acause2

% cause is something hich has an effect' 4n epidemiolog! a cause can be considered

to be something that alters the fre3uenc! ofdisease, health status or associated factorsin a population'

Pragmatic definition' Philosophers hae grappled ith the nature

of causalit! for thousands of !ears'

7/24/2019 18481


Some philosoph! Daid Hume5s philosoph! has been influential'

% cause cannot be deduced logicall! from the fact that toeents are lin)ed'

Because thunder follos lightning does not mean thunderis caused b! lightning' *bsering this one million timesdoes not ma)e it true'

+he a-iom 6%ssociation does not mean causation7' Cause and effect deductions need more than obseration

alone . the! need understanding' +he contribution of another philosopher, John Stuart 0ill,

captured in his canons, is so similar to the modernempiricall! based ideas of epidemiolog!'

7/24/2019 18481


"pidemiological strateg! and

reasoning: the e-ample of Semeleis Diseases form patterns, hich are eer changing'Clues to the causes of disease are inherent ithin

these pattern'Semeleis 89#9#.9#;< obsered that the

mortalit! from childbed feer 8no )non aspuerperal feer< as loer in omen attendingclinic = run b! midies than it as in thoseattending clinic 9 run b! doctors'

Do these obserations spar) off an! ideas ofcausation in !our mind2

7/24/2019 18481


Births, deaths, and mortalit! rates 8>< for

all patients at the to clinics 9#?9.9#?

7/24/2019 18481


Semmeleis@ inspiration

4n 9#?A, his colleague and friend Professorolletsch)a died folloing a fingerpric) ith a )nifeused to conduct an autops!'

olletsch)a@s autops! shoed inflammation to beidespread, ith peritonitis, and meningitis'

“Day and night I was haunted by the image ofKolletschka’s disease and was forced to recogniseever more decisively that the disease from whichKolletschka died was identical to that from which somany maternity patients died '7

Semeleis5 inspired idea as that particles had beentransferred from the scalpel to the ascular s!stem ofhis friend and that the same particles ere )illingmaternit! patients'

7/24/2019 18481


Semmeleis@ action

4f so, something stronger than ordinar!

soap as needed for handashing

He introduced chlorina li3uida, and thenfor reasons of econom!, chlorinated lime'

+he maternal mortalit! rate plummeted'

Semeleis@s discoer! as resented in

ienna'

.

7/24/2019 18481


essons from Semmeleis@s or) Deep )noledge deries from the e-planation of disease

patterns, rather than in their description'

4nspiration is needed, and ma! come from une-pectedsources, as here from olletsch)a@s autops!' %ction cannot ala!s aait understanding the mechanism' "pidemiological data to sho that la!ing an infant on its

front 8prone position< to sleep raises the ris) of 5cot death5

or sudden infant death s!ndrome' % campaign to persuade parents to la! their infants on theirbac)s has haled the incidence of cot death'

"pidemiologists are reliant on other sciences, laborator! orsocial, to be e3ual partners, in pursuit of the mechanisms'

7/24/2019 18481


"pidemiological principles and

models of cause and effect

0ost important of the cause and effect ideas underpinned b!epidemiolog! is that disease is irtuall! always a result of theinterpla! of the enironment, the genetic and ph!sical ma)eupof the indiidual, and the agent of disease'

Diseases attributed to single causes are inariabl! so b!

definition' +he fact that 6tuberculosis7 is 6caused7 b! the tuberclebacillus is a matter of definition'

+he causes of tuberculosis, from an epidemiological orpublic.health perspectie, are man!, including malnutritionand oercroding'

+his idea is captured b! seeral ell )non disease causationmodels, such as the line, triangle, the heel, and the eb'

7/24/2019 18481


1igure ;'=

Clues

Stable in incidence

Clusters in families

Clues

4ncidence aries rapidl!

oer time or beteengeneticall! similar

populations

4s the disease predominantl!

genetic or enironmental2

&enetic

"nironmental

7/24/2019 18481


1igure ;'E

Down’s syndrome

Phenylketonuria

Sickle cell disease

Diabetes

Asthma

Coronary heart disease

Stroke Lung cancer

Road traffic

accidents

&enetic "nironmental

7/24/2019 18481


1igure ;'?

Host

"nironment%gent

+he underl!ing cause of the

disease is a result of the

interaction of seeral factors,

hich can be anal!sed using

the components of the

epidemiological triangle'

7/24/2019 18481


1igure ;';

Host:Inhalation of infectie

organism! age! smoking!male se"!

cardio#res$iratory disease

"nironment:Presence of cooling towers

and com$le" hot water

systems% aerosols created

but not contained!

meteorological conditions

take aerosol to humans

%gent:

&irulentLegionella

organisms! e.g.

$neumo$hila

seroty$e

7/24/2019 18481


1igure ;'

Control smo)ing

and causes of

immunodeficienc!

%oid et t!pe cooling

toers, loo) for a betterdesign and location,

separate toers from

population and enhance

toer h!giene

0inimise groth of

organisms and factorshich enhance

pathogenicit!, e'g' algae

7/24/2019 18481


Ph!sicalenironment

Socialenironment

Chemical Fbiological

enironment

!ene

host

1igure ;'A

+he model

emphasises the unit!

of the gene and host

ithin an interactie

enironmental

enelope

+he oerlap beteen

enironmental

components

emphasises the

arbitrar! distinctions

7/24/2019 18481


1igure ;'#

Ph!sical

enironment:

aailability of

health care

facilities fordiagnosis

Social

enironment:

social su$$ort

to sustain

dietary change

Chemical F

biological

enironment:

diet content

!ene defect

en#yme

deficiencybrain

damage

7/24/2019 18481


0odels of cause and effect %gent factors, arguabl!, receie less attention

than the! desere' Characterising the irulence of organisms is difficult' 4n other diseases conceptualising the cause as an agent

is not eas!' +he concept of the disease agent has been applied to

infections but it or)s ell ith man! non.infectiousagents, for e-ample, cigarettes, motor cars, and alcohol' +he interaction of the host, agent and enironment is

rarel! understood' +he effect of cigarette smo)ing is substantiall! greater in

poor people than in rich people'

7/24/2019 18481


0odels of cause and effect

"ach model is a simplification'0oe from simple to comple- models' +he categories of host, agent and

enironment are arbitrar!' +he host and agent are, of course, both

part of the enironment'

"nironment, in this conte-t, is arbitraril!defined to mean factors e-ternal to thehost and the agent of disease'

7/24/2019 18481


+he triangle and preention

+he epidemiological triangle can be

combined ith the schema of theleels of preention to deise a

comprehensie frameor) for

thin)ing about possible preentie

actions'

7/24/2019 18481


0odels: the heel

+he heel of causation' "mphasises the unit! of the interacting factors' "mphasises the fact that the diision of the

enironment into components is somehat arbitrar!'

0odel is applied to phen!l)etonuria, the archet!palgenetic disorder'

Phen!l)etonuria is an autosomal single genedisease '

%n enG!me re3uired to metabolise the dietar! amino.acid phen!lalanine and turn it into t!rosine, isdeficient'

7/24/2019 18481


+he heel: phen!l)etonuria

Brain damage is the outcome' +he cause of this disease could be said to be a

gene'

+he cause of the disease could be consideredas a combination of a gene'

"-posure to a chemical and biologicalenironment hich proides a diet containing a

high amount of phen!lalanine' % social enironment unable to protect the child

from the conse3uences, of a gene disorder'

7/24/2019 18481


0odels: the spider@s eb

1or man! disorders our understanding of thecauses is highl! comple-' "ither the causes are trul! comple-, or e3uall! li)el!,

our understanding is too rudimentar! to permit clarit!' +hese disorders are referred to as multifactorial or

pol!factorial disorders' 0echanisms of causation are not apparent' Portra!ed b! the metaphor of the spider@s eb' +his modelindicates the potential for the disease to

influence the causes and not just the other a! around,

so.called, reerse causalit!' 4t also poses a fundamental 3uestion: here is the

spider that spun the eb2

7/24/2019 18481


7/24/2019 18481


7/24/2019 18481


4ndiidual e-ercise on

geneIenironment interaction

+hin) about a disease that one of !our friends or relatieshae had'''e-cept for those e hae discussed

Reflect on the causes using the line, triangle and heel ofcausation'

%t !our leisure: +hin) through the cause of disease K using these models

8bo- 9', chapter 9<' 4s disease K li)el! to be genetic or enironmental2 h!2

&o oer !our ansers ith !our classmates

7/24/2019 18481


%nal!sing diseases using the

heel and eb models

Reie the health problems or diseases

that !ou pic)ed and disease K 8Chapter 9,

bo- 9'< using the heel and eb models'

L d ffi i t

7/24/2019 18481


Lecessar! and sufficient cause ast5s Dictionar! tells us that a necessar! cause is M%

causal factor hose presence is re3uired for theoccurrence of the effect7 , and,

Sufficient cause as a 6minimum set of conditions, factorsor eents needed to produce a gien outcome7' +he tubercle bacillus is re3uired to cause tuberculosis but,

alone, does not ala!s cause it, so it is a necessar!, not asufficient, cause'

Consider the causes of Don@s s!ndrome 8+risom! =9<,sic)le cell disease, tuberculosis, scur!, phen!l)etonuria,and lung cancer'

hen a specific cause of disease is sufficientl! ell )non

it can be incorporated into its definition 8as in Don5s

S!ndrome, sic)le cell disease and itamin C deficienc!<'

R th @ t d l

7/24/2019 18481


Rothman@s component causes model Rothman5s interacting component causes model has

emphasised that the causes of disease comprise aconstellation of factors'

4t has broadened the sufficient cause concept to be a minimalset of conditions hich together ineitabl! produce thedisease'

+he concept is shon in figure 99 +hree combinations of factors 8%BC, B"D, %C"< are shon

here as sufficient causes of the disease' "ach of the constituents of the causal MpieM are necessar!' Control of the disease could be achieed b! remoing one of

the components in each MpieM and if there ere a factorcommon to all MpiesM the disease ould be eliminated b!remoing that alone'

7/24/2019 18481


1igure ;'99

% B

C

% "

D

% "

C

"ach of the three components of theinteracting constellations of causes

8%BC, %D", %C"< are in themseles

sufficient and each is necessar!

& id li f id i l i l

7/24/2019 18481


&uidelines for epidemiological

reasoning on cause and effect +urning epidemiological data into an understanding of

cause and effect is challenging' "pidemiologists need an e-plicit mode of reasoning' Subjectie judgements on cause and effect in

epidemiolog! should not be dismissed'

"pidemiologists place much more emphasis on theealuation of empirical data'

Criteria for causalit! proide a a! of reaching judgements on the li)elihood of an association beingcausal'

% frameor) for thought, applied before ma)ing a judgement, based on all the eidence'

" id i l i l it i

7/24/2019 18481


"pidemiological criteria

8guidelines< for causalit!

Causal criteria in microbiolog!, health economics,philosoph! offer much to epidemiolog!'

Henle.och postulates' 0ill@s canons

"conomics also ealuates associations in similar a!s' %ccording to CharemGa and Deadman, the operational

meaning of causalit! in economics is more on the linesof 5to predict5 than 5to produce5 8an effect<'

"pidemiological criteria are, hoeer, designed forthin)ing about the causes of disease in populationsand not in indiiduals'

"pidemiological thin)ing in cause

7/24/2019 18481


"pidemiological thin)ing in cause

and effect "pidemiolog! establishes causes in populations

but this information applies to indiiduals in a probabilistica!' hich does not proe cause and effect at the indiidual leel ' 4f $N> of all lung cancer in a population is due to smo)ing,

hat is the li)elihood that in an indiidual ith lung cancer thecause as smo)ing2

+here is no a! to distinguish a lung cancer resulting fromsmo)ing from a lung cancer arising from another cause'

% factor demonstrated to cause a disease in an indiidual, sa!using to-icolog! or patholog!, ma! not be demonstrable asharmful in the population' h!2

imitation of a science of indiiduals'

7/24/2019 18481


%pplication of guidelinesIcriteria

to associations

%n association rarel! reflects a causal

relationship but it ma!'

+hese si- criteria are a distillation of, or at

least, echo the ten %lfred "ans5

postulates in ast5s Dictionar! of

"pidemiolog! 8?th edition< and the nine

Bradford Hill criteria'

7/24/2019 18481


+emporalit!

Did the cause precede the effect2 4f the effect follos the action of a proposed cause

the association ma! be a causal one and theanal!sis can proceed'

+hunder follos lightning' Does lightning cause

thunder2 4f !ou flic) a sitch and a light goes on, can !ou

deduce that !ou and !our action cause the light togo on2

Just because B follos %, does not of itself, confirma causal relation' Deeper understanding or openingthe blac) bo- is essential'

7/24/2019 18481


Strength and dose response

Does e-posure to the cause change disease

incidence2 4f not there is no epidemiological basis for a conclusion oncause and effect'

1ailure to demonstrate this does not, hoeer, disproe acausal role'

+he usual measure of the increase in incidence is the relatieris) and the technical name for this criterion is the strengthof the association'

Dose.response Does the disease incidence ar! ith the leel of e-posure2

4f !es, the case for causalit! is adanced' +he dose.response relation is also measured using the

relatie ris)'

7/24/2019 18481


Specificit!

4s the effect of the supposed cause specific toreleant diseases, and, are diseases caused b! alimited number of supposed causes2

4magine a factor hich as lin)ed to all health

effects h! ould that be so2 Lon.specificit! is characteristic of spurious

associations eg underestimating the siGe of thedenominator'

hile specificit! is not a criticall! important criterionepidemiologists should ta)e adantage of thereasoning poer it offers'

7/24/2019 18481


Consistenc!

4s the eidence ithin and beteen

studies consistent2

Consistenc! is lin)ed to generalisabilit! of

findings'

Spurious associations are often local'

7/24/2019 18481


"-periment

Does changing e-posure to the supposed

cause change disease incidence2

*ften there hae been natural e-periments'

Deliberate e-perimentation ill be necessar!'

Human e-periments or trials are sometimes

impossible on ethical grounds'

Causal understanding can be greatl! adancedb! laborator! and e-perimental obserations'

7/24/2019 18481


Biological plausibilit!

4s there a biological mechanism b! hich thesupposed cause can induce the effect2

1or trul! noel adances, hoeer, thebiological plausibilit! ma! not be apparent'

Biologicall! plausible that la!ing an infant on itsbac) to sleep ma! lead to its inhaling omitus'

*erturned b! the biologicall! implausibleobseration that la!ing a child on its bac)

hales the ris) of cot death' Lonetheless, biological plausibilit! remains

releant to establishing causalit!'

Judging the causal basis of the

7/24/2019 18481


Judging the causal basis of the

association +he criteria are particularl! aluable in e-posing the

lac) of eidence for causalit!, for indicating the needfor further research and for aoiding prematureconclusions'

Sometimes firm judgements are possible' Sometimes, judgments are forced upon us' +hree e-amples of the case for causalit! in boo)' Dieth!lstilboestrol as a cause of adenocarcinoma of

the agina 8Herbst et al<' Smo)ing as a cause of lung cancer, 8Doll et al< and

Residential pro-imit! to a co)ing or)s as a cause ofill.health 8Bhopal et al<'

"-ample of judging causalit!: lung cancer

7/24/2019 18481


"-ample of judging causalit!: lung cancer

causalit!: lung cancer

7/24/2019 18481


causalit!: lung cancer

1i ; 9E +h id f i ti

7/24/2019 18481


1igure ;'9E +he p!ramid of associations

9 Causal and mechanisms

understood

= Causal

E Lon.causal

? Confounded

; Spurious I artefact

Chance

4 t t ti f d t t d

7/24/2019 18481


4nterpretation of data, stud!

design and causal criteria

Causal )noledge is born in the imagination andunderstanding of the disease process of theinestigator'

Same data can be interpreted in 3uite different a!s' +he paradigm ithin hich epidemiologists or) ill

determine the nature of the causal lin)s the! see andemphasise'

Researchers to ma)e e-plicit in their ritings theirguiding research philosoph!'

Lo epidemiological design confirms causalit! and nodesign is incapable of adding important eidence'

7/24/2019 18481


1igure ;'9= +he scales of causal judgement

'eigh u$ weaknesses in data

and alternatie e"$lanations

'eigh u$ (uality of science

and results of a$$lying causal

frameworks

" id i l i l th

7/24/2019 18481


"pidemiological theor!

illustrated b! this chapter

Diseases arise from a comple- interactionof genetic and enironmental factors'

Causes of disease in indiiduals ma! not

necessaril! be demonstrable causes ofdisease in populations and ice ersa'

Cause and effect judgements areachieable through h!pothesis generationand testing, ith data interpreted using alogical frameor) of anal!sis'

7/24/2019 18481


Summar!

Cause and effect understanding is the highestform of scientific )noledge'

"pidemiological and other forms of causal

thin)ing shos similarit!' %n association beteen disease and the

postulated causal factors lies at the core ofepidemiolog!'

Demonstrating causalit! is difficult because ofthe comple-it! and long natural histor! ofman! human diseases and because of ethicalrestraints on human e-perimentation'

7/24/2019 18481


Summar!

%ll judgements of cause and effect aretentatie'

Be alert for error, the pla! of chance and

bias'Causal models broaden causal perspecties'%ppl! criteria for causalit! as an aid to

thin)ing' oo) for corroboration of causalit! from

other scientific frameor)s'

18481

Documents