On the Origin of Hill's Causal Criteria

0 downloads 0 Views 589KB Size Report
On the Origin of Hill's Causal Criteria ... the 1964 Report of the Advisory Committee to the US. Surgeon ... inferences and that the discovery of causes required an.
On the Origin of Hill's Causal Criteria Author(s): Alfredo Morabia Source: Epidemiology, Vol. 2, No. 5 (Sep., 1991), pp. 367-369 Published by: Lippincott Williams & Wilkins Stable URL: http://www.jstor.org/stable/20065702 Accessed: 06/01/2010 11:15 Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=lww. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Lippincott Williams & Wilkins is collaborating with JSTOR to digitize, preserve and extend access to Epidemiology.

http://www.jstor.org

On the Origin of Hill's Causal Criteria Alfredo Morabia

The

to assess

rules

Hill's

Bradford

Hume's

of whether

causation work

was

"The

as opposed

D. Hume,

causality,

Environment

lack of

The

contemporary epidemiologists. and Hill's, analogy of Hume's Keywords:

A.B.

research

on

association.

As

recently

the noted

in

Here I compare the causal criteria formulated by Hume and by Hill. A first contrast is of historical nature. In

of

1740,

or Causa

reference nature

causal

in a review

by

Susser,2 Hill codified in his paper a set of criteria that had since 1955.3"6 been elaborated by several epidemiologists Part of these criteria was applied in assessing the causal relation

between

cigarette

and

smoking

cancer

lung

in

to the US the 1964 Report of the Advisory Committee Surgeon General, Smoking and Health.1 The rationale was that a finding satisfying several criteria was more likely to be causal than one that satisfied only a few or none. While reading David Hume's A Treatise of Human Nature,8 I came upon the "rules by which to judge of causes and effects" (pp. 173-176) and was struck by their to HilPs

similarities proaches

to

criteria. To in

causality

find

contemporary

and in an eighteenth century philosopher raised the question of the philosophic criteria that, to my knowledge, has not heretofore. In addition, a potential link

analogous

ap

epidemiologists

was puzzling. It roots of Hill's been addressed between Hume

and Hill

(read: between Hume and the generation of epidemiologists who contributed to the elaboration of the criteria efficiently summarized by Hill) suggested that there could be a parallelism between the Popper-Hume and the Popper-Hill controversies. The first controversy is

now

controversy, generated

to

well-known

initiated

a variety

epidemiologists.9

The

to Hill's

criteria,

from rejection of the criteria reflecting inductive rejection of the method itself.12

From the Clinical Micheli-du-Crest

second

15 years ago by Carol Buck,10 has

of criticisms

1991;2;367?369)

(Epidemiology

K. Popper.

is a classic

assessing

a point of view still widely shared by expresses thinking to causal in epidemiology inferences the may explain

proof

experimental logic.

Association

tion," by Sir A. Bradford Hill,1 epidemiologic an observed

systematic

to Popper's,

Hill,

Disease:

and

are compared to Sir Austin David Hume philosopher rules and Hill's causal criteria that, irrespective suggests

Scottish century by the eighteenth Hume's of the analogy between strength to Hill or Hill's Hume's known predecessors, formulated

The

criteria.

causal

Epidemiology Unit, University Canton 25, 1211 Geneva 4, Switzerland.

ranging

logic11 to

Hospital,

mental

?

1991 Epidemiology

Primus

Resources

Inc.

for assistance

in the preparation

was

method

not

the

but

emerging, a system

yet

of

experi

research.

Hume

believed that observations were fragile bases for causal inferences and that the discovery of causes required an intellectual process going "beyond the impression of our senses."

He

therefore

a process.

such

a set of

proposed

In contrast,

Hill

and

to structure

rules his

contemporary

epidemiologist colleagues, concerned by the biases result ing from the "method of observation" in studying human populations,13 tried to define the circumstances in which we

can

infer

causation

from

observa

nonexperimental

tion.

Methods I have tried to match Hill's are

in

expressed

and Hume's two

their

criteria as they Because

publications.1'8

treatise was published 225 years before Hill's it is obviously impossible to get a perfect match. report,

Hume's For

example,

Hume

his

presented

as

rules

statements, whereas Hill's criteria are worded for preventive

medicine.

My

goal,

light on a potential philosophical the formulation itself.

universal

specifically was

however,

to

shed

kinship rather than on

Results The

of Hume's

comparison

rized inTable 1. The conceptual called

and Hill's

is striking for what Hill has

identity

"temporality,"

is summa

criteria

"biologic

gradient,"

and

"consis

tency." The temporality criterion is stated explicitly by both Hume and Hill. Hill: "The temporal relationship of is the cart and which the association?which is the Rue

horse?" in space

The author thanks Maryline the manuscript.

were

sciences

empirical

of

Hume: and

"The time.

The

cause

and

cause

(Rules land 2). The concept of gradient associations is also mentioned

effect

must

must

be prior

be

contiguous

to the

effect"

as a support for causal by both authors, but for

367

MORABIA

1. Hill's Criteria* and Corresponding Hume's

TABLE

"Rules

to Judge

by which

of Causes

mu?a

and

Effects"t Hill's

Hume's

Criteria

Rules cause must

1. "The

2. Dose-response

2.

3. Consistency

3.

of association

be prior

to

criterion

cause"

7. Biological plausibility

7. Not applicable4 8. Not

Experiment

tion.

we

then

curve,

...

evidence.

should

The

clear

or diminishes

its

to

'tis

cause,

with

be

a

as

regarded

dose-response

and obviously puts "When any object

increase

the

most

look

or diminution

some

"In

Hill:

rule.

viral

Hume's

example, and

effect,

circum

in pregnancy."

disease

(Rule 4)

always

never

effect

echoes

produces

arises

but

from

criterion

Hill's

of

reasons, two remaining criteria listed by have no plausibility and experiment) set

in Hume's

of

It would

rules.

an

be

anachronism to find the concept of biological plausibility in the writing of a seventeenth century philosopher. Also, although a contemporary of Newton, Hume (1711-1776) was a preexperimental thinker. Experimental method became

a

century,

after

of

system

research

the works

and

others.

agreed with Hume's

causation,

the

emphasized

without

cannot

but

Indeed,

approach on

of experiment: can ascertain

role

experimentation

co-existences,

nineteenth

the

during

of Lavoisier

John Stuart Mill, who

and

effect,

cause

"same same

the

same cause"

tion

of

compounded

sixth

or another

drug

counterpart

curve admits of a simple explanation the case in a clearer light." Hume: increases

Hume's

specificity. For historical Hill (biological

applicable.^:

Hume this is a sufficient causal criterion. Hill: "If the association is one which can reveal a biological gradient, carefully

For same

the

1. *Source: Reference f Source: Reference 8, pp. 173-176. $See Results section.

such

constant

Hume: "Where several different objects produce the same effect, itmust be by means of some quality, which we discover to be common amongst them" (Rule 5). "Like effects imply like causes" (Rule 6). It is possible to find in Hume's writing formulas that correspond to Hill's criteria of specificity and coherence.

the

4).

(Rule

8.

for

a

be

the importance lies in the similarity of the Again, intellectual approach rather than in the exact formula

causes" (Rule 6). "Same cause always produces the same effect, and the same effect never arises but from

6.

dose-response

of association

must

it would

another

with

rule but in the premises of the p. 163). catalog, "There must be a constant union betwixt the cause and

the same

or

a measure

risk, "There

be fair to judge by analogy. With the effects of thalidomide and rubella before us we should surely be ready to accept slighter but similar evidence

effect" (Rule 3). 5. "Like effects imply like

Specificity

and

stances

increases "When any object or diminishes with the or diminution increase of its cause" (Rule 7). of resembling "Multiplicity instances constitutes the very essence of power or connexion" (not a specific

4.

5. Analogy 6.

relative

betwixt the cause and effect. 'Tis chiefly this quality, that constitutes the relation" (Rule 3). is some resemblance between Hill's analogy There

the effect" (Rules 1 and 2).

Strength

the

causation:

of

than

union

1. Temporality

4.

like

is, just

rather

"Observa sequences

causation."14

prove

deriv'd from the union of the several different effects, which arise from the several different parts of the cause. The

of

criterion

second

proves

sufficiently

conjunction

is the cause of the other" Hill's

part

cause

the

is here

to be always attended with the absence or of a proportionable part of the effect. This

suppos'd presence constant

of one

or presence

absence

the one

that

instances

to whether

[that]

of "multiplicity

concept constitutes

the

very

"the

that

or

of power

(not a specific rule but in the premises of the catalog, p. 163). Strength effect, rules.

does

not

Nevertheless,

the

association,

have

an Hume's

exact

as

a measure

complement constant-conjunction

of

expresses

Hume's

criteria

between For

course

in Hume's

proof

of

and

inferences of nature

must will

rely not

on

vary.

the Since

it

is

the debate

illuminate

and non-Popperian is no

us,

sounded reasonable epidemiologists?

causation.

Cause-to

effect relations can be inferred from past experience, such

relative

predecessors,3"7

to

reasonable

sound

this conclusion there

whether

a point of view still widely The point is epidemiologists.

criteria would have

Popperian

Hume,

or Hill's

and Hill's of

irrespective

to Hill

known

thinking

likely that Hill's to Hume. Does

of resembling

essence

that,

suggests was

shared by contemporary

associa

connexion"

of

work

Hume's

tion [has] been repeatedly observed by different persons, in different places, circumstances and time?" The coun terpart is Hume's

criteria

causal Hume's

part

(Rule 7). relates

Discussion The strength of the analogy between Hume's

that

assumption there

is no

proof

but the that

causally related events will still be related in the future in

for

368

Epidemiology

September

1991, Volume

2 Number

5

ON THE ORIGIN OF HILL'SCAUSAL CRITERIA

same

the

can

effect

the

ways, never

of an

validity

from

inference

cause

to

be demonstrated.15

Popper has argued against Hume that the scientific community could temporarily consider as valid theories that had resisted repeated tests. By tests, Popper means can

that

experiments

a matter

As

anyone.16

be

principle?by

reproduced?in

of

is historical

there

fact,

is Newton's

laws

evidence

Epidemiologists are

ments

of motion.17

may

agree with to

tools

important

that experi

Popper

identify

causal

Lind

laws.

was a pioneer of the experimental method.18 Textbooks of refer to Lind, Louis,19 and so on20 to epidemiology roots of experiments in epidemio historical the highlight can hardly share research. But, logic epidemiologists theories are thesis that testable central only Popper's deals with

scientific.21 Epidemiology human

and

populations

the characteristics is more

therefore

an

of

observa

tional than an experimental discipline.13 Epidemiologists often have to infer causation without being able to bring objective

For

proofs. was

connection to controlled

the

example,

elucidated

although

it was

not

amenable

experimentation.

I suggest that Popper's philosophy appears too restric tive for epidemiologists because it does not recognize the scientific character of theories that do not have potential experimental proof. Hume's philosophy that proofs are elusive

is also

by most

rejected

scientists,

contemporary

proof is not available, pragmatic epidemiolo gists simply acknowledge that there is yet no alternative

but when to the

causal

criteria

inferring causation, ties

and

rule

out

logic

that

says,

in substance:

before

it is imperative to check for illogicali

gross

contradictions

between

what

has

found and what we think we know. Hence, the as to Hume's and of Hill's, analogy Popper's, opposed

been

logic.

References 1. Hill AB. The environment and disease: Association Proc R Soc Med 1965;58:295-300.

Epidemiology

September

1991, Volume

or causation?

2 Number

AM. On in chronic

the methodology Some diseases.

of etiologic of investigations comments. Dis J Chronic

1959;10:41-46. of etiologic the methodology of investigations 6. Sartwell PE. On Dis Further comments. factors in chronic diseases. J Chronic 1960;11:61-63. and Health. Report of the Advisory Committee 7. Smoking No. Public Health Service Publication Surgeon General. of Health, DC: US Department Washington 1964. Welfare, of Human Nature. Second 8. Hume D. A Treatise Oxford

Press, University 9. Rothman KJ, ed. Causal

to the 1103.

Education,

and

edition. Oxford:

1978.

Inferences. Chestnut Hill: Epidemiologie Inc. 1988. Resource, Int ] Epidemiol 10. Buck C. Popper's philosophy for epidemiologists. 1975;4:159-168. 11. Weed DL. Causal criteria and Popperian In: Rothman refutation. Resource, KJ, ed. Causal Inferences. Chestnut Hill: Epidemiologie Inc. 1988; 15-32. 12. Lanes Causal

SF. The

In: Rothman logic of causal inferences. KJ, ed. Hill: Epidemiologie Inferences. Chestnut Inc. Resource,

1988;59-76. 13. Hill AB. Observation

cancer

smoking-lung

5. Lilienfeld factors

that such successful theories are likely to remain valid within the conditions of their discovery. A classic exam ple

is a cause and how do we know one? A grammar for 1991;133:635-648. pragmatic epidemiology. Am J Epidemiol 3. Hammond EC. Cause and Effect. In:Wynder ES, ed. The Biologic Effect of Tobacco. Boston: Little, Brown and Co, 1955; 171-196. of investigations of 4. Yerushalmy J, Palmer CE. On the methodology etiologic factors in chronic diseases. JChronic Dis 1959;10:27-40. 2. Susser M. What

5 369

and experiment. New Engl JMed 1953;248: 995-1,001. and Inductive. 1843; 243. 14. Mill JS. A System of Logic, Ratiocinative in Holland Cited PW. Statistics inferences. and causal JASA 1986;81:945-960. 15. Hume D. An Abstract In: Hume of A Treatise of Human Nature. D. An

Human Understanding. Enquiry Concerning Court, 1988;27-43. 16. Popper KR. The Logic of Scientific Discovery. New and Row, 1968;44-45.

La Salle,

IL:

Open

York: Harper

A, Infeld L. The Evolution of Physics. New York: Touch stone, 1966;237. 18. Lind J.An inquiry into the nature, causes and cure of the scurvy. In: Buck C, Llopis A, N?jera E, Terris M, eds. The Challenge of Issues and Selected Readings. Washington DC: Pan Epidemiology. American Health Organization, 1988; 20-23. on the 19. Lilienfeld AM. The French influence DE, Lilienfeld 17. Einstein

of epidemiology. In: Lilienfeld AM, ed. Times, Places of the History of Epidemiology. Baltimore: The Johns Hopkins University Press, 1980;28-38. 20. Buck C, LLopis A, N?jera E, Terris M, eds. The Challenge of development and Persons.

Aspects

Issues and Selected Readings. Washington Epidemiology. American Health Organization, 1988. 21. Miller D, ed. Popper Selections. Princeton: Princeton Press, 1985; 122-123.

DC:

Pan

University