Unit 6 Causal Inference_1000e6d4-93d6-4826-b237-2d3b0263ac5d_2ff828fe-122f-46b3-b11b-dcac6dcb0fde.pptx

BHCM SEM IV Notes

1. Unit 6: Causal Inference

2. Outline: Define the terms cause and causal inference Differences between association and causation Sufficient-component cause model Hill's criteria for causality and the limitations of these criteria

3. Meaning of cause and causality Cause It is an event, condition, or characteristics which play an important role or predicable change in occurrence of the outcome Example: smoking and lung cancer Causality (causation / cause-effect relationship): -it is the term that describes relating causes to the effects they produce.

4. Cause: A cause is something that produces an effect. Webster’s dictionary defines a cause as ‘something which has an effect’. In epidemiology a cause can be considered to be something that alters the frequency of disease, health & production status or associated factors in a population. An event, condition, characteristic or a combination which plays an important role in occurrence of the outcome. E.g. smoking and lung cancer

5. Factors involved in causation Precipitating factors – exposure to agent Predisposing factors – age,sex,previous illness Enabling factors – malnutrition, low socioeconomic status Reinforcing factors- repeated exposure

6. Causal inference: Causal inference in epidemiology involves determining the causal relationship between an exposure and an outcome in a population. This is often done through observational studies, where researchers analyze data to infer causality while considering potential confounding variables and biases. Techniques such as propensity score matching, instrumental variable analysis, and sensitivity analysis are commonly used to strengthen causal inference in epidemiological studies. Fig: Causal relationship

7. Cont …. A causal inference is a statement about why something happens . It states the existence of a relationship between at least two variables. The dependent variable measures that variation which we would like to explain (find a cause for). Also called Y, or the “outcome” or “response” or variable. The independent variable measures that variation which we think explains variation in the dependent variable. Also called X, or the “treatment” or “study” variable. The causal effect of a treatment is the difference between what happens to a unit after that treatment and what would have happened had the unit not been treated.

8. The field of causal inference consists of 3 parts: 1 A formal language for unambiguously defining causal concepts . This is just a formalization of the common sense we already have. 2 Causal diagrams: a tool for clearly displaying causal assumptions . They can be used to inform both the design and analysis of observational studies. 3 Analysis methods (i.e. statistical methods ) that can help us draw more reliable causal conclusions from the data at hand.

9. Types of causal relationship:

10. A factor is both necessary and sufficient for producing the disease Without that factor, the disease never develops & in the presence of that factor, the disease always occurs. 1. Necessary and Sufficient ‘A’ No ‘A’ No Disease Disease

11. Each factor is necessary, but not in itself sufficient to cause the disease Multifactor are required, often in a specific temporal sequence 2. Necessary, but not Sufficient Factor 1 Factor 2 Factor 3 Reaction at Cellular Level Disease

12. 3. Sufficient, but not Necessary Factor 1 Factor 2 Factor 3 Reaction at Cellular Level Disease

13. 4. N e ither Sufficient nor Necessary Factor A+B Factor C +D Factor E+F Reaction at Cellular Level Disease OR OR

14. Cont …..

15. Cont …..

16. Association and causation

17. Association and causation

18. ASSOCIATION: Association may be defined as the concurrence of two variables more often than would be expected by chance. In other words, events are said to be associated when they occur more frequently together than one would expect by chance. Association does not necessarily imply a causal relationship. The terms "association" and "relationship" are often used interchangeably. In epidemiology, we determine the relationship/association between given exposure and frequency of the disease in a population. Association is a statistical relationship between the exposure and disease.

19. Types of association: Association can be broadly grouped under three headings : 1. Spurious association 2. Indirect association 3. Causal association (direct association) a) One-to-one causal association b) Multifactorial causation.

20. 1. Spurious association Sometimes an observed association between a disease and suspected factor may not be real. For example, a study in UK of 5174 births at home and 11,156 births in hospitals showed perinatal mortality rates of 5.4 per 1000 in the home births, and 27.8 per 1000 in the hospital births. Apparently, the perinatal mortality was higher in hospital births than in the home births. It might be concluded that homes are a safer place for delivery of births than hospitals..

21. Cont …. Such a conclusion is spurious or artifactual, because in general, hospitals attract women at high risk for delivery because of their special equipment and expertise, whereas this is not the case with home deliveries. The high perinatal mortality rate in hospitals might be due to this fact alone, and not because the quality of care was inferior. There might be other factors also such as differences in age, parity, prenatal care, home circumstances, general health.

22. Example of spurious association

23. 2. Indirect association Many associations which at first appeared to be causal have been found on further study to be due to indirect association. The indirect association is a statistical association between a characteristic (or variable) of interest and a disease due to the presence of another factor, known or unknown, that is common to both the characteristic and the disease. The third factor is also known as confounding variable.

24. 3. Direct (causal) association (a) One-to-one causal relationship Two variables are stated to be causally related (AB) if a change in A is followed by a change in B. If it does not, then their relationship cannot be causal. This is known as "one-to- one" causal relationship.

25. Cont …. This model suggests that when the factor A is present, the disease B must result. Conversely, when the disease is present, the factor must also be present.

26. Example of Direct Causation

27. Cont ………. (b) Multifactorial causation The causal thinking is different when we consider a non-communicable disease or condition (e.g., CHO) where the etiology is multifactorial.

28. Association and Causation

29. Summary:

30. Models of causality • Sufficient-component cause model (Rothman’s pies) • Counterfactual framework • Hill’s criteria or “viewpoints” • Graphical model • Causal perspective on effect modification

31. SUFFICIENT-COMPONENT CAUSE MODEL

32. Sufficient-component cause model In 1976 Ken Rothman, proposed a conceptual model of causation known as the "sufficient-component cause model" in an attempt to provide a practical view of causation which also had a sound theoretical basis. The model has similarities to the "web of causation" theory, but is more developed in the sense that it simultaneously provides a general model for the conditions necessary to cause (and prevent) disease in a single individual and for the epidemiological study of the causes of disease among groups of individuals.

33. A Sufficient Cause: Rothman recognized that disease outcomes have multiple contributing determinants that may act together to produce a given instance of disease. For example: exposure to someone who has Tuberculosis (TB) does not necessarily result in the occurrence of TB. Moreover, the set of determinants that produce TB in one individual may not be the same set of conditions that were responsible for the occurrence of TB in others.

34. Cont ….. Rothman defined a sufficient cause as a complete causal mechanism that inevitably produces disease. Consequently, a "sufficient cause" is not a single factor, but a minimum set of factors and circumstances that, if present in a given individual, will produce the disease . Aschengrau and Seage used the example of causation of AIDS.

35. A sufficient cause for AIDS might consist of the following components: exposure to an individual with HIV repeatedly engaging in risky sexual behavior with that individual absence of antiretroviral drugs that reduce viral load of HIV

36. Cont … The model suggests that the presence of these three component causes is sufficient to produce AIDS in this individual. ( Note further if any one of these components were absent, AIDS would not occur.) Hence, Rothman's assertion that a cause is an event, condition, or characteristic without which the disease would not have occurred. Note that the sufficient cause illustrated here is only one manner in which AIDS could occur. Different individuals will have different sets of individual components that combine to produce a sufficient cause (i.e., a case of AIDS).

37. If one were to apply the sufficient-component cause model to tuberculosis (TB), one possible cause might be represented by the pie chart given here:

38. This sufficient cause may have applied to many of the people who developed TB in the United Kingdom during the 19th and 20th century. The line graph below shows the annual mortality from TB per 100,000 population from 1860 to 1950.

39. Cont … During this time span the introduction of "the hygienic idea" and the subsequent development of public health initiatives led to gradual improvements in living conditions, including less crowding, better ventilation, and better nutrition. The decreased prevalence of these components is likely to have been responsible for the steady decline in TB mortality seen during this period. Note, however, the two points on the line graph that correspond to World War I and World War II when there are temporary increases in TB mortality. It is well known that the wars had a widespread impact on the population and that nutrition suffered and people were sometimes seeking shelter in bomb shelters that were poorly ventilated and crowded.

40. Cont …. The sufficient-component model to the left offers a coherent explanation for the cause of TB mortality in a large proportion of the population during this period, and it also explains the steady decline punctuated with the temporary increases seen during war time. However there may be many sufficient causes of TB which may differ in their components, although some components might be shared among different sufficient causes.

41. Consider, for example, the two sufficient causes below.

42. Cont …… Among the three sufficient causes of TB illustrated above, there are both similarities and differences in the composition of the components. They also differ in the number of components. For example, an individual with AIDS and poor nutrition would be severely immune-compromised, so the only component needed to complete the causation of TB would be exposure to the TB bacillus.

43. Features of the Sufficient-Component Cause Model Aschengrau and Seage point out some of the key features of the sufficient-component cause model: A cause is not a single component, but a minimal set of conditions or events that inevitably produces the outcome. Each component in a sufficient cause is called a component cause , and epidemiologists tend to refer to the components as "causes" because the outcome will not occur by that pathway if any one of the components is missing (or prevented) within a given sufficient cause model. Consequently, it is not necessary to identify all of the component causes in order to prevent the disease outcome.

44. Cont… 3. There may be a number of sufficient causes for a given disease or outcome. A component cause that must be present in every sufficient cause of a given outcome is referred to as a necessary cause . For example: HIV exposure is necessary for AIDS to occur TB exposure is necessary for TB infection to occur .

45. Cont… 5. The completion of a sufficient cause is synonymous with the biologic occurrence of the outcome, e.g., the transition to a malignant cancer within a single cell marks the biologic onset of the cancer. 6. The components of a sufficient cause do not need to act simultaneously ; they can act at different times. For example, a mutation in a proto-oncogene in a prostate cell may promote cell replication at one point in time, and it may be some time later when another mutation diminishes the function of an anti-oncogene in the same cell.

46. Cont ….. Thus, each component cause may have a different induction period text annotation indicator (the interval between the exposure's presence and disease onset). In contrast, the latent period is the interval between disease onset and the clinical detection of disease, either by screening or as a result of symptoms and diagnostic work up. In the context of screening tests the latent period is referred to as the "detectable pre-clinical phase." In the context of infectious disease, it is the time between initial infection and the first appearance of symptoms.

47. The Sufficient Cause Model Lots of ways to get a disease Think of each way as a pie Called a sufficient cause Mechanisms exist independent of us But we’re susceptible to them if we acquire the components Go through life picking up exposures and filling in pies 56

48. The Sufficient Cause Model 57 Person is susceptible to multiple diseases Diseases have multiple sufficient causes Each sufficient cause has multiple component causes Each component cause has attributes Shared components between sufficient causes It is theoretically possible every case of outcome has a unique pie

49. HILL'S CRITERIA FOR CAUSALITY

50. Hill's criteria for causality The  Bradford Hill criteria , otherwise known as  Hill's criteria for causation , are a group of 9 principles, established in 1965 by the English epidemiologist Sir  Austin Bradford Hill . They can be useful in establishing  epidemiologic  evidence of a  causal relationship between a presumed cause and an observed effect and have been widely used in  public health  research. Their exact application and limits of the criteria continue to be debated.

51. Cont…. In 1965, the epidemiologist, Austin Bradford Hill, who helped  link smoking to lung cancer ,  gave a speech  where he presented his  viewpoints  on how we can arrive at causation from correlation. This lecture was a bit of a game changer at the time given that the tobacco industry was employing statisticians, medical doctors, and  even popular science writers  to push the idea that the relationship between smoking and lung cancer was merely a correlation, not a causal one.

52. Hill’s Criteria for Causal Association Bradford Hill recognized the importance of moving from association to causation as a necessary step for taking preventive action against environmental causes of disease. Bradford Hill’s criteria for making causal inferences: Temporal association Strength of association Dose-Response relationship Lack of temporal ambiguity Consistency of findings Biologic plausibility Coherence of evidence Specificity of association Consideration of alternate explanation

53. 1. Temporal association (progressive association) Does the suspected cause precede the observed effect? A causal association requires that exposure to a putative (supposition) cause must precede temporarily the onset of a disease which it is purported (assumed) to produce to allow for any necessary period of induction and latency . Lung cancer occurs in smokers of long-standing; this satisfies the temporal requirement. Further, the increase in consumption of cigarettes preceded by about 30 years the increase in death rates from lung cancer . Smoking( Cause) Precedes Cancer (Effect)

54. 2.Strength of association The stronger the association, the more likely it is that the relation is causal. With increasing level of exposure to the risk factor an increase in incidence of the disease is found. This can be calculated either by Odds ratio or Relative Risk Relative risk is the ratio of the incidence of the disease among exposed and the incidence among non-exposed.

55. For Example- Risk for development of lung cancer is 8.6 times higher in smokers than in non-smokers.

56. Relative Risk Relative Risk = Incidence among Exposed Incidence among Non Exposed RR = 1 No association RR > 1 Positive association (possibly causal) RR < 1 Negative association (possibly protective) It is direct measure of the strength of association.

57. Odds Ratio Cases(with disease) Controls (without disease) History of exposure a b No history of exposure c d Odds Ratio = OR = ad/ bc Odds of disease in exposed group Odds of disease in Non-exposed group

58. 3.Dose-Response relationship As the dose of exposure increases, the risk of disease also increases. If present, it is strong evidence for a causal relationship. Presence of D-R relationship strengthens Causality, whereas its absence doesn’t rule out Causal relationship. Absence of a dose-response relationship does not necessarily rule out a causal relationship. • In some cases in which a threshold may exist, no disease may develop up to a certain level.

59. Example: Daily average cigarettes smoked Relative risk of developing lung cancer 1 - 14 6.7 15-24 12.3 25+ 23.7

60. 4.Lack of temporal ambiguity (vague/unclear) • Ambiguous temporal  precedence is lack  of clarity about which variable occurred first may yield confusion about which variable is the cause and which is the effect. Exposure to the factor must have occurred before the disease developed • The temporal relationship is important in regard to the length of the interval between exposure and disease. • It’s easier to establish a temporal relationship in a prospective cohort study than in a case-control study or a Retrospective cohort study.

61. Example- Consumption of contaminated food should precede the symptoms of food poisoning.

62. 5.Consistency of findings • The relationship should be found consistently in different studies and in different populations . • Unless there is a clear reason to expect different results, replication of the findings should be there. If the relationship is causal, we would expect to find it consistently in different studies and in different populations. Causal Association betweenn Smoking and Lung cancer is found consistently in: 50 retrospective studies, 9 prospective studies.

63. 6.Biologic plausibility • The association must be consistent with the current knowledge of disease. ( viz mechanism of action, evidence from animal experiments etc ). Sometimes the lack of plausibility may simply be due to the lack of sufficient knowledge about the pathogenesis of a disease. Biologic plausibility refers to coherence with the current body of biologic knowledge. Epidemiologic findings should be consistent with existing biologic knowledge. Example- Carcinogens from cigarette smoke deposits in the lung over a period of time leading to lung cancer.

64. 7. Coherence of evidence • The association must be coherent with the known facts of relevant origins. If a relationship is causal, we would expect the findings to be consistent with other data. • For the appraisal of causal significance of an association it should be coherent with known facts that are thought to be relevant.

65. Example - 1. Male and Female differences in trends of lung cancer. Deaths is coherent with recent adoption of Cigarette smoking by women. 2. Peptic ulcer disease • Prevalence of H.pylori is same inn men as in women. Incidence of duodenal ulcer in both have been proved to be equal in recent years. • Prevalence of peptic ulcer disease is believed to have peaked in the latter part of 19 th century cause of poor living standards.

66. Coherence of the Association

67. 8. Specificity of association • Specificity implies a one to one relationship between the cause and effect (Weakest Criteria). Association is specific when a certain exposure is associated with only one disease • When specificity of an association is found, it provides additional support for a causal inference • Absence of specificity in no way negates a causal relationship.

68. Cont ……. Example 1- Prevalence of H.pylori in patients with duodenal ulcer is 90% to 100%. However, it is found even in some patients of gastric ulcer and even in asymptomatic individuals. Example 2- Not everyone who smokes develops lung cancer. Not everyone who develops cancer has smoked.

69. 9. Consideration of alternate explanations In judging whether a reported association is causal, the extent to which the investigators have taken other possible explanations into account and the extent to which they have ruled out such explanations are important considerations. Deriving/growing Causal inferences by eliminating- Bias, Confounding and Chance etc.

70. Limitations of Hill’s criteria There’s no completely reliable means of establishing a causal relationship and sometimes evidence can be conflicting. To make a causal inference, all available evidence must be considered. Correct Temporal relationship is very essential before other criteria are considered ( plausibility, consistency and dose-response relationship ). The likelihood of a causal association is heightened when many different types of evidence lead to the same conclusion.

71. Limitations of Hills criteria 1. Strength of association A strong association is more likely to have a causal component that is a modest association Hill illustrated this point with the high risk ratios for the association between exposure levels of smoking and incidence of lung cancer. However, he demonstrated with two counter-examples that the absence of a strong association does not rule out a causal effect. Hill acknowledged that the impression of strength of association depended on the index used for the magnitude of association.   2. Consistency A relationship is observed repeatedly. For Hill, the repeated observation of an association included "different persons, places, circumstances and time". The benefit of this rule was that consistently finding an association with different study designs (e.g. in both retrospective and prospective studies) reduced the probability that an association would be due to a "constant error or fallacy" in the same study design. On the other hand, he pointed out that shared flaws in different studies would tend to replicate the same wrong conclusion. Likewise, differing results in different investigations might indicate that some studies correctly showed a causal relationship, whereas others failed to identify it.

72. 3. Specificity A factor influences specifically a particular outcome or population. For Hill, if one observed an association that was specific for an outcome or group of individuals, this was a strong argument for a causal effect. In the absence of specificity, Hill alludes to fallacies in applying this rule to conclude the absence of a causal effect: Diseases may have more than one cause (which Hill considered to be the predominant case). In turn, a factor might cause several diseases. According to Hill, the value of this rule lay in its combination with the strength of an association: For instance, among smokers, the risk of death from lung cancer should be elevated to a higher degree as compared to the risk of other causes of death. Hill's consideration on specificity for persons apparently contradicts his consideration on consistency, where repeatedly observing an association in different populations would increase the evidence for a causal effect.

73. 4. Temporality [timing of information about cause and effect] The factor must precede the outcome it is assumed to affect Hill introduced this reflection with the proverb "Which is the cart and which is the horse?" For instance, he asked whether a particular diet triggered a certain disease or whether the disease led to subsequently altered dietary habits. According to Hill, temporal direction might be difficult to establish if a disease developed slowly and initial forms of disease were difficult to measure.   5. Biological gradient [dose-response relationship] The outcome increases monotonically with increasing dose of exposure or according to a function predicted by a substantive theory Hill favoured linear relationships between exposure level and outcome, for instance, between the number of cigarettes smoked per day and the death rate from cancer. If the shape of the dose-response relationship were a more complex, especially a non-monotonic, function, this would require a more complex substantive explanation.

74. 6. Plausibility [putative(common) cause] The observed association can be plausibly explained by substantive matter (e.g. biological) explanations For Hill, the presence of a biological explanation supported the drawing of a causal conclusion. On the other hand, in the absence of such a theory "the association we observe may be one new to science or medicine and we must not dismiss it too light-heartedly as just too odd".   7. Coherence [logical] A causal conclusion should not fundamentally contradict present substantive knowledge. Hill used the term "generally known facts" to indicate that the knowledge against which an association is evaluated has to be undisputable. Laboratory evidence that is in line with an association would underline a causal conclusion and help to identify the causal agent. Again, the absence of such knowledge would not be indicative of a non-causal explanation .

75. 8. Experiment Causation is more likely if evidence is based on randomized experiments , Hill argued that a causal interpretation of an association from a non-experimental study was supported if a randomized prevention derived from the association confirmed the finding. For instance, after finding that certain events were related to the number of people smoking, one might forbid smoking to see whether the frequency of the events decrease consecutively.   9. Analogy [compare] For analogous exposures and outcomes an effect has already been shown Hill wrote that it would be sometimes acceptable to "judge by analogy". He gives the following example: "With the effects of thalamoide and rubella before us we would surely be ready to accept slighter but similar evidence with another drug or another viral disease in pregnancy ."

76. ASSIGNMENT-6 (Short questions) Define cause and causal inference. What is causal association? What are the types of causal relationships? How can association be grouped into? (types of association) Difference between association and causation. List Hills criteria for causation. List the Limitations of Hill’s criteria. Define risk factor.

77. ASSIGNMENT-6 (Long Questions) Write briefly about the types of causal relationships. Explain the types of association between variables. What is causal association in epidemiology? Explain Bradford Hills criteria and also list limitation of Hills criteria. Explain the types of causal relationship. Explain the Rothmans pie concept of disease causation. (Sufficient component cause model)