Causal Inference Dr Amna Rehana Siddiqui Department of

Causal Inference Dr. Amna Rehana Siddiqui Department of Family and Community Medicine

Objectives: n Explain basic models of disease causation. n To understand concepts related to scientific inference for cause effect relation n To understand the applicability of causal criteria as applied to epidemiological studies

Approach to etiology n To see whether a certain substance is an agent / microorganism; a controlled laboratory experiment can be done by ¨ Exposing animals to organism ¨ Setting the exposure dose ¨ Monitoring environmental conditions ¨ Selecting genetic factors ¨ Minimum loss to follow up ¨ Species differ in response

Observations in Human populations n n Cannot randomize human beings for harmful substances Depend on nonrandomized observations Important populations – occupational cohorts Natural experiments ¨ Residents of Hiroshima and Nagasaki ¨ Residents of Bhopal

Stages of disease and Levels of prevention Primary prevention Susceptibility n n Pre-symptomatic n Secondary (Screening) n Clinical n Tertiary prevention n Disability or Recovery n Tertiary prevention n prevention

Development of Disease Combination of events n n n A harmful agent A susceptible host An appropriate environment

General Models of Causation n In epidemiology, there are several models of disease causation that help understand disease process. n The most widely applied models are: ¨ The epidemiological triad (triangle), ¨ the wheel, and ¨ the web. And ¨ The sufficient cause and component causes models (Rothman’s component causes model) 7

The Epidemiologic Triad HOST AGENT 8 ENVIRONMENT

Agent factors • Infectious agents: agent might be microorganism—virus, bacterium, parasite, or other microbes. e. g. polio, measles, malaria, tuberculosis Generally, these agents must be present for disease to occur. • Nutritive: excesses or deficiencies (Cholesterol, vitamins, proteins) • Chemical agents: (carbon monoxide, drugs, medications) • Physical agents (Ionizing radiation, … 9

Host factors • Host factors are intrinsic factors that influence an individual’s exposure, susceptibility, or response to a causative agent. • Host factors that affect a person’s risk of exposure to an agent: • e. g. Age, race, sex, socioeconomic status, and behaviors (smoking, drug abuse, lifestyle, sexual practices and eating habits) • Host factors which affect susceptibility &response to an agent: • Age, genetic composition, nutritional and immunologic status, anatomic structure, presence of disease or medications, and psychological makeup. 10

Environmental factors are extrinsic factors which affect the agent and the opportunity for exposure. Environmental factors include: ¨ physical factors such as geology, climate, . . ¨ biologic factors such as insects that transmit an agent; and ¨ socioeconomic factors such as crowding, sanitation, and the availability of health services. 11

Malaria Agent Vector 12 Host Environment

The epidemiologic triad Model Host: Intrinsic factors, genetic, physiologic factors, psychological factors, immunity Health or Illness ? Agent: 13 Amount, infectivity, pathogenicity, virulence, chemical composition, cell reproduction Environment: Physical, biological, social

Web of Causation n There is no single cause n Causes of disease are interacting n Illustrates the interconnectedness of possible causes 14

The Web of causation Developed to de-emphasis agent n Chain of causation n Complexity of origin is web n Multiple factors promote or inhibit n Emphasizes multiple interactions between host and environment n

zat ion Web of Causation en o org ani ph typ soc ial e microb es r be Disease envir fa n w nk no U ce pla 16 rk wo ct o rs genes ou i v ha onme nt

Web of Causation - CHD sc ep ns atio su dic tib me ne s stres ge ili lipids ty g n i ok sm Disease physi c al act ivity fa n w nk no U n tio ma lam inf 17 blood pressure ct o rs er gend RS Bhopal

Example of a Web of Causation Overcrowding Malnutrition Exposure to Mycobacterium Susceptible Host Infection Tuberculosis Tissue Invasion and Reaction Vaccination 18 Genetic

The Wheel of Causation n The Wheel of Causation de-emphasizes the agent as the sole cause of disease, n It emphasizes the interplay of physical, biological and social environments. It also brings genetics into the mix. 19

The Wheel of Causation Social Environment Biological Environment Host (human) Genetic Core 20 Physical Environment

Association Vs. Causation n Association refers to the statistical dependence between two variables n The presence of an association…in no way implies that the observed relationship is one of cause and effect 21

Types of causes n Sufficient causes: ¨a set of conditions without any one of which the disease would not have occurred ¨ not usually a single factor, often several n Necessary cause: ¨ must 22 be present for disease to occur, disease never develops in the absence of that factor. ¨ a component cause that is a member of every sufficient cause

The sufficient cause and component causes model Rothman’s component causes model 23

Necessary and sufficient causes n A necessary cause is a causal factor whose presence is required for the occurrence of the effect. If disease does not develop without the factor being present, then we term the causative factor “necessary”. n Sufficient cause is a “minimum set of conditions, factors or events needed to produce a given outcome. n The factors or conditions that form a sufficient cause are called component causes. 24

Example n The tubercle bacillus is required to cause tuberculosis but, alone, does not always cause it, n so tubercle bacillus is a necessary, not a sufficient, cause. 25

Rothman’s Component Causes and Causal Pies Model n Rothman's model has emphasised that the causes of disease comprise a collection of factors. n These factors represent pieces of a pie, the whole pie (combinations of factors) are the sufficient causes for a disease. n It shows that a disease may have more that one sufficient cause, with each sufficient cause being composed of several factors. 26

Rothman’s Component Causes and Causal Pies n The factors represented by the pieces of the pie in this model are called component causes. component n Each single component cause is rarely a sufficient cause by sufficient itself, But may be necessary cause. necessary n Control of the disease could be achieved by removing one of the components in each "pie" and if there were a factor common to all "pies“ (necessary cause) the disease would be eliminated by removing that alone. 27

Exercise n Some of the risk factors for heart disease are smoking, hypertension, obesity, diabetes, high cholesterol, inactivity, stress, and type A personality. - Are these risk factors necessary causes, sufficient causes, or component causes? 28

Causal pies representing all sufficient causes of a particular disease 29

Types of Associations Real: probability depends upon the occurrence of one or more other events, characteristics, or other variables n Spurious: Non causal associations depend on bias, chance, failure to control for extraneous variables (confounding) n

Percentage of pregnancies (n=50, 267) with infant weighing < 2500 g by maternal cigarette smoking category (peri-natal mort study Comm Vol 1, 1967 % less than 2500 g

Percentage of LBW infants by smoking status of their mothers (Yerushalmay J, Am J Obs & Gynecol) % of LBW infants Non Smoker All pregnancies Future All Preg Future Smoker Ex smoker

“Is there an association between an exposure and a disease? ” IF SO…. n Is the association likely to be due to chance? n Is the association likely to be due to bias? n Is the association likely to be due to confounding? n Is the association real/causal? 33

Establishing the cause of disease Association? absent Chance? present Bias ? likely Confounding? likely present absent 34 Causal?

Association Vs. Causation n Association refers to the statistical dependence between two variables n The presence of an association…in no way implies that the observed relationship is one of cause and effect 35

Epidemiological criteria (guidelines) for causality n An association rarely reflects a causal relationship but it may. n Criteria for causality provide a way of reaching judgements on the likelihood of an association being causal. 36

Hill’s Criteria for Causal Relation n Strength of association n Consistency of findings n Specificity of association n Temporal sequence n Biological gradient (dose-response) n Biological plausibility n Coherence with established facts n Experimental evidence 37

Strength of association n Does exposure to the cause change disease incidence? n The strength of the association is measured by the relative risk. n The stronger the association, the higher the likelihood of a causal relationship. n Strong associations are less likely to be caused by chance or bias 38

Consistency of findings n Consistency refers to the repeated observation of an association in different populations under different circumstances. n Causality is more likely when the association is repeated by other investigations conducted by different persons in different places, circumstances and time-frames, and using different research designs. 39

Specificity of association n It means that an exposure leads to a single or characteristic effect, or affects people with a specific susceptibility ¨ easier to support causation when associations are specific, but ¨ this may not always be the case n n as many exposures cause multiple diseases This is more feasible in infectious diseases than in noninfectious diseases, which can result from different risk agents. 40

Temporal sequence (temporality) n Did the cause precede the effect? n Temporality refers to the necessity that the cause must precede the disease in time. n This is the only absolutely essential criterion. n It is easier to establish temporality in experimental and cohort studies than in case-control and cross-sectional studies. 41

Biological gradient n Does the disease incidence vary with the level of exposure? (dose-response relationship) n Changes in exposure are related to a trend in relative risk n A dose-response relationship (if present) can increase the likelihood of a causal association. 42

Biological gradient (Dose Response) 43

Age standardized death rates due to bronchogenic carcinoma by current amount of smoking n 44 Dose-response relationship

Biological plausibility n Is there a logical mechanism by which the supposed cause can induce the effect? n Findings should not disagree with established understanding of biological processes. 45

Coherence n Coherence implies that a cause-and-effect interpretation for an association n does not conflict with what is known of the natural history and biology of the disease 46

Experimental evidence n It refers to evidence from laboratory experiments on animal or to evidence from human experiments n Causal understanding can be greatly advanced by laboratory and experimental observations. 47

Judging the causal basis of the association n No single study is sufficient for causal inference n It is always necessary to consider multiple alternate explanations before making conclusions about the causal relationship between any two items under investigation. n Causal inference is not a simple process ¨ consider weight of evidence ¨ requires judgment and interpretation 48

Figure 5. 12 The scales of causal judgement Weigh up weaknesses in data and alternative explanations 49 Weigh up quality of science and results of applying causal frameworks

Pyramid of Associations Causal Non-causal Confounded Spurious / artefact Chance 50 RS Bhopal

Evaluating Evidence of Causal relationship Major Criteria a. Temporal relationship b. Biologic plausibility c. Consistency of Results d. Alternative explanations Other criteria a. Strength of association b. Dose-response relationship c. Cessation of effects n 51