LESSON 1.4 WORKBOOK - Tufts University · module, this type of interventional study is the only way...

8
27 Workbook Lesson 1.4 Notes: ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ ________________________________ So how can we determine whether something “causes" cancer? In 1947 the British Ministry of Health became aware that lung cancer deaths had risen nearly 15-fold compared to the previous two decades. It seemed that an unprecedented epidemic was in progress. In response, the Medical Research Council, the body responsible for handling government-funded scientific research in Britain, organized a conference of experts to investigate and hopefully find a cause for this unexpected rise. The conference decided to establish a thorough study, run by the statistician Austin Bradford Hill, to identify risk factors for lung cancer. Hill was given a shoestring budget, which forced him to hire inexperienced staff such as Richard Doll, who, though he was a medical researcher, had never organized a study of such scale or significance before. LESSON 1.4 WORKBOOK How can we identify carcinogens? In order to fully understand cancer and to develop effective treatments we need to know how it is caused. This lesson compares Koch’s postulates – the criteria that must be satisfied in order to estab- lish an agent as infectious, with Hill’s postulates, the criteria that must be satisfied to establish an agent as carcinogenic. But how can we definitively prove that a carcinogen causes cancer in humans if we clearly can’t ethically cause the disease? This problem results in certain ambiguities about what is a carcinogen and what is not that companies have exploited to avoid liability.

Transcript of LESSON 1.4 WORKBOOK - Tufts University · module, this type of interventional study is the only way...

Page 1: LESSON 1.4 WORKBOOK - Tufts University · module, this type of interventional study is the only way to conclusively prove causation, but that ... Le 1.4 LESSON READINGS MC Questions:

27

W o r k b o o kLesson 1.4

Notes: ________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

So how can we determine whether something “causes" cancer?

In 1947 the British Ministry of Health became aware that lung cancer deaths had risen nearly 15-fold compared to the previous two decades. It seemed that an unprecedented epidemic was in progress. In response, the Medical Research Council, the body responsible for handling government-funded scientific research in Britain, organized a conference of experts to investigate and hopefully find a cause for this unexpected rise. The conference decided to establish a thorough study, run by the statistician Austin Bradford Hill, to identify risk factors for lung cancer. Hill was given a shoestring budget, which forced him to hire inexperienced staff such as Richard Doll, who, though he was a medical researcher, had never organized a study of such scale or significance before.

LESSON 1.4 WORKBOOKHow can we identify carcinogens?

In order to fully understand cancer and to develop effective treatments we need to know how it is caused. This lesson compares Koch’s postulates – the criteria that must be satisfied in order to estab-lish an agent as infectious, with Hill’s postulates, the criteria that must be satisfied to establish an agent as carcinogenic. But how can we definitively prove that a carcinogen causes cancer in humans if we clearly can’t ethically cause the disease? This problem results in certain ambiguities about what is a carcinogen and what is not that companies have exploited to avoid liability.

Page 2: LESSON 1.4 WORKBOOK - Tufts University · module, this type of interventional study is the only way to conclusively prove causation, but that ... Le 1.4 LESSON READINGS MC Questions:

LESSON READINGS

28

W o r k b o o kLesson 1.4

DEFINITIONS OF TERMS

Case-control study – a study where patients with a disease and

patients without disease are re-viewed to identify patterns that may have led to development of disease.

Cases – the group within a case-control study that developed

disease.

Controls ¬– the group within a case-control study that did not

develop disease.

Over-reported – when survey re-spondents claim they partook of an activity more than they actually did.

Under-reported – when survey re-spondents claim they partook of an activity less than they actually did.

Interventional study – a study where one group is given a treat-ment to evaluate the effect of that

treatment on people.

MC Questions:

1. True or false: A case-control study involves providing a treatment to a 'case' group and not to a 'control' group.

.a True.

.b False.________________________________________________________________________________________________

2. Why can’t Koch’s postulates be applied to finding causes of cancer? (Circle all correct.)

.a You can’t identify patterns of exposure for carcinogens and cancer.

.b You cannot inject carcinogens into humans to cause cancer.

.c Few cancers are caused by pathogens.

.d You cannot isolate carcinogens from cancer patients.

________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

Doll and Hill worked together to design a case-control study with the goal of figuring out what the lung cancer patients had in common. They interviewed patients admitted to one of 20 London hospitals for either lung cancer (cases) or other illnesses (controls), about various aspects of their life, including how close their homes were to gasworks, how often they ate fried fish, whether they ate bacon, sausage or ham for dinner, and their smoking habits. Doll and Hill hoped these questions would identify factors that led one group to developing lung cancer, while the other group remained healthy.

Doll and Hill’s study (and another by a group in the US) showed the same result in two different populations, in two different countries. Even though the studies used different methodologies they both arrived at the same conclusion: smoking predisposes people to lung cancer. Perhaps more astonishing, the magnitude of risk in both studies was almost identical, proving the strength of the association of smoking to lung cancer.

Despite the seemingly convincing results, Hill was very concerned that his study might be biased. In particular he was worried that the extent of smoking was over-reported in lung cancer victims and under-reported in the control group. If this was the case, the relevance of smoking in lung cancer would be skewed. He realized that the best way to conclusively prove a link would be to randomly assign patients to two different groups one of which would smoke while the other would not. He would then follow how frequently members of each group would develop cancer. As we saw in the Metabolic Diseases module, this type of interventional study is the only way to conclusively prove causation, but that ghoulish human experiment would violate many ethical principles in medicine.

So, Hill did the next best thing, which was to simply observe a population of people that contained both smokers (treatment group) and nonsmokers (control group) for 20 months and then determine the frequency of lung cancer deaths in each group. There were 40,000 people in the study, and 789 died over the 20-month period, 36 due to lung cancer. Of those 36, every single one was a smoker! The relationship between smoking and lung cancer was beyond statistical.

Figure 1: Design of a case-control experiment. By examining the histories of two distinct populations, one can identify risk factors that led one group to acquire disease while the other group remained healthy.

Page 3: LESSON 1.4 WORKBOOK - Tufts University · module, this type of interventional study is the only way to conclusively prove causation, but that ... Le 1.4 LESSON READINGS MC Questions:

LESSON READINGS

29

W o r k b o o kLesson 1.4

MC Questions:

3. True or False: All 9 of Hill’s postulates must be met in order to show a relationship is causal.

.a True.

.b False.

________________________________________________________________________________________________

4. Why do many correlations not show causation? (Circle all correct.)

.a There may be hidden variables that explain the true causation.

.b Correlation data comes from poorly done experiments.

.c Causation can only be shown by Koch’s postulates.

.d Correlation data is only one line of evidence. Up to 8 other criteria are necessary to show causation.

________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

DEFINITIONS OF TERMS

Hill’s Postulates – a list of epidemiological criteria that set

the minimal evidence necessary to show causality.

For a complete list of defined terms, see the Glossary.

When these studies were eventually published in 1956, tobacco companies quickly realized the implications and began a tightly orchestrated disinformation campaign. They published a newspaper advertisement stating: ‘recent reports on experiments with mice have given wide publicity to a theory that cigarette smoking is in some way linked with lung cancer in human beings’ a clear mis-statement of facts that, by tying Hill’s data to mice rather than humans, made the link appear less relevant. Tobacco companies also accused scientists of conflating a correlation between smoking and lung cancer into a causal relationship.

Causation vs. Correlation: Hill’s postulates

This battle was not only over a specific link between smoking and lung cancer, it was over the merits of the emerging field of epidemiology that had started with Percival Pott and scrotal cancer. As we learned in the Infectious Disease module, the only way that we can actually prove that a pathogen causes an infectious disease is first to isolate it and then to show that the isolated pathogen causes disease in a susceptible host. Rous used the exactly the same principle to show his RSV caused chicken sarcoma. However Rous was fortunate in two ways we are often not: first, unlike most carcinogens his virus would cause the tumor after only one exposure. Second, Rous was not working with human cancer. As Bradford Hill also realized, it's simply not ethical to try to identify a potential human carcinogen by inducing to produce cancer in humans. This left the cancer field in a quandary. Epidemiologists perform observations that can at best only demonstrate strong correlations, and Koch’s postulates would not apply in a complex disease such as cancer where multiple mutations might be needed for cancer to develop. Were we ever going to prove causation?

Grappling with that question, and taking his cue from Koch and his postulates, Bradford Hill came up with a list of 9 distinct criteria, that, if they were all met, would provide very strong evidence supporting causality for any chronic and complex human disease. Though Hill realized that the only way to prove causality is with Koch-type studies that were neither feasible in humans, nor appropriate for complex diseases, he argued that his postulates were like a detective’s case, in which many small pieces of evidence can substitute for a single convincing piece of evidence.

Figure 2: Design of an interventional study. Groups are randomly assigned to an interven-tion/treatment or control group. Effects of that intervention are measured after some time, and control and intervention groups are compared.

Page 4: LESSON 1.4 WORKBOOK - Tufts University · module, this type of interventional study is the only way to conclusively prove causation, but that ... Le 1.4 LESSON READINGS MC Questions:

LESSON READINGS

30

W o r k b o o kLesson 1.4

DEFINITIONS OF TERMS

Underlying variables – also known as “hidden” variables, are

factors that obscure a relationship that is believed to be causal.

Notes: ________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

Hill’s Postulates

When applied to smoking and lung cancer, Hill’s postulates strongly support causality, as seen below:

1. Plausibility — the link can be explained by biological explanations

2. Strength — a strong link is more likely to be causal than a weak one

3. Consistency — the link is observed repeatedly in different settings

4. Specificity — the link is a single factor causing a specific disease

5. Temporality — the factor exposure must precede onset of disease

6. Biological gradient — the extent of exposure should be proportional to risk of developing disease

7. Coherence — the relationship does not disagree with accepted scientific knowledge

8. Experiment — randomized intervention experiments in model animals produces disease

9. Analogy — a similar effect should be seen in other exposed populations/organs

As we mentioned in the Metabolic Disease module, many correlations do not prove causation. For any correlation, there are underlying, or “hidden” variables that may explain the relationship. For instance, studies show that eating food coloring is correlated to hyperactivity, but that does not mean it is causative. Foods that have food coloring typically may also have sugar, which is also correlated to hyperactivity, and would be a hidden variable.

Figure 3: Koch’s postulates were the first way of showing that an agent is responsible for causing disease. Unfortunately, Koch’s postulates are limited to diseases caused by infectious agents, and can not be used for cancer.

Page 5: LESSON 1.4 WORKBOOK - Tufts University · module, this type of interventional study is the only way to conclusively prove causation, but that ... Le 1.4 LESSON READINGS MC Questions:

31

W o r k b o o kLesson 1.4

LESSON READINGSMC Questions:

5. True or false: Lung cancer is caused by genetics as well as smoking, so it is not preventable.

.a True.

.b False.________________________________________________________________________________________________

6. Why is it hard to prove something causes cancer? (Circle all correct.)

.a It is difficult to remove all hidden variables.

.b Companies have financial interests in selling cancer-causing products.

.c People are unlikely to believe something that is popular will cause cancer.

.d Scientists are not good at communicating data to public.

________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

DEFINITIONS OF TERMS

Meaningful relationship – a type of relationship that cannot be shown to be causal, but has attempted to remove as many hidden variables as possible.

The challenge of identifying causative agents in cancer is that there is no ethical way to do an interventional study on carcinogens causing cancer in people. Therefore, Hill’s postulates is a good substitute for identifying causal relationships by identifying meaningful relationships. The structure of the postulates is that by meeting more criteria, you remove alternative explanations. If for instance, scientists gave mice food coloring agents (without sugar) and found that they caused cancer, this would support the relationship between food coloring and hyperactivity. The more Hill’s postulates that are met, the more likely a relationship is meaningful.

This history of how the association between smoking and lung cancer was established also reveals that lung cancer is a preventable disease. We can dramatically reduce our risk of developing lung cancer by not smoking. It is worth mentioning that a drop in lung cancer incidence occurred following bans on smoking in public places like bars and restaurants that became commonplace during the 1990’s. This approach towards removing the carcinogens in tobacco from our society is a good model for limiting exposure to other carcinogens as a tool to prevent cancer.

Why is it so hard to prove something 'causes' cancer?

Let’s say hypothetically, that there is a contaminant in tobacco that causes cancer, there is no way of separating that contaminant’s role in causing cancer from tobacco’s role in causing cancer. Nevertheless, all tobacco products are shown by Hill’s postulates to be associated with cancer. Therefore, the relationship is not causal, but meaningful.

Tobacco companies took advantage of this inability to prove causality in the 50's and 60's by releasing ads distorting the evidence of a link between smoking and cancer. The relationship of smoking and lung cancer was still in dispute for many people until the 1980s, where legal action against tobacco companies forced them to reveal internal documents that proved that not only did tobacco companies know there was a link between their product and cancer, but also that they suppressed research results demonstrating the addictive properties of nicotine. The American public finally began to pay attention to the overwhelming

Figure 4: The time lag of 20 years between cigarette consumption and onset of lung cancer makes it difficult to see the relationship between cigarettes and lung cancer.

Page 6: LESSON 1.4 WORKBOOK - Tufts University · module, this type of interventional study is the only way to conclusively prove causation, but that ... Le 1.4 LESSON READINGS MC Questions:

32

W o r k b o o kLesson 1.4

LESSON READINGSMC Questions:

7. Which of the following helped reduce the incidence of lung cancer in the US? (Circle all correct.)

.a Banning smoking from public areas.

.b Campaigns to raise awareness of smoking and cancer.

.c Tobacco companies making safer cigarettes.

.d Increased consumption of e-cigarettes.

8. Why do people think e-cigarettes may be hazardous to health? (Circle all correct.)

.a They meet all 9 Hill’s postulates for cancer.

.b They are smoked by the elderly.

.c They contain known carcinogens.

.d They are made from tobacco.________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________

DEFINITIONS OF TERMS

E-cigarette – an electronic cigarette that is advertised as a “healthy” alternative to smoking,

but involves inhaling a vapor containing formaldehyde and

other known carcinogens.

scientific evidence, and the legal wrangling led to government restrictions on tobacco advertising and taxes that collectively decreased the average consumption of cigarettes from 4,141 cigarettes per person in the US in 1974 to 2,500 cigarettes consumed per person in the US in 1994.

After these setbacks, the tobacco industry re-focused their efforts toward the developing world and into non-cigarette tobacco usage. At the present time, 50% of men and 11% of women in developing nations smoke cigarettes, and those numbers are increasing. Meanwhile, in the US, per capita cigarette consumption continues to decrease from 2,076 in 2000 to 1,232 in 2011 (a 32.8% decrease), but per capita cigar and loose leaf tobacco consumption has increased from 72 in 2000 to 142 in 2011 (a 96.9% increase).

Consumption of electronic cigarettes (or e-cigarettes), also known as 'vaping' because it involves inhalation of a chemical vapor, is advertised as a healthy alternative to smoking. However, the vaporized liquid solution of e-cigarettes contains formalde-hyde, a known carcinogen. Consumption of e-cigarettes has increased exponentially, and sales of e-cigarettes is expected to exceed $10 billion by 2020. There is insufficient data on the risk of consuming e-cigarettes to lung or mouth cancers, so Hill’s postulates cannot be determined for the relationship between e-cigarettes and cancer. Neverthe-less, there is concern for the consumption of e-cigarettes, as they are unregulated by the government, and children as young as 10 are consuming these products.

Figure 5: The time lag of 20 years between cigarette consumption and onset of lung cancer makes it difficult to see the relationship between cigarettes and lung cancer.

Page 7: LESSON 1.4 WORKBOOK - Tufts University · module, this type of interventional study is the only way to conclusively prove causation, but that ... Le 1.4 LESSON READINGS MC Questions:

STUDENT RESPONSES

33

W o r k b o o kLesson 1.4

Remember to identify your sources

Look up online an agent that is said to 'cause' cancer. Based upon the article, how many postulates are met from the data presented? Do you think this agent causes cancer? Why or why not?

_____________________________________________________________________________________________________

_____________________________________________________________________________________________________

_____________________________________________________________________________________________________

_____________________________________________________________________________________________________

_____________________________________________________________________________________________________

_____________________________________________________________________________________________________

____________________________________________________________________________________________________

_____________________________________________________________________________________________________

_____________________________________________________________________________________________________

_____________________________________________________________________________________________________

_____________________________________________________________________________________________________

_____________________________________________________________________________________________________

_____________________________________________________________________________________________________

_____________________________________________________________________________________________________

_____________________________________________________________________________________________________

_____________________________________________________________________________________________________

_____________________________________________________________________________________________________

_____________________________________________________________________________________________________

_____________________________________________________________________________________________________

_____________________________________________________________________________________________________

_____________________________________________________________________________________________________

_____________________________________________________________________________________________________

_____________________________________________________________________________________________________

_____________________________________________________________________________________________________

___________________________________________________________________________________________

Page 8: LESSON 1.4 WORKBOOK - Tufts University · module, this type of interventional study is the only way to conclusively prove causation, but that ... Le 1.4 LESSON READINGS MC Questions:

34

W o r k b o o kLesson 1.4

TERMS

For a complete list of defined terms, see the Glossary.

TERM DEFINITION

Cases The group within a case-control study that developed disease.

Case-control study A study where patients with a disease and patients without disease are reviewed to identify patterns that may have led to development of disease.

Controls The group within a case-control study that did not develop disease.

E-cigarette An electronic cigarette that is advertised as a “healthy” alternative to smoking, but involves inhaling a vapor containing formaldehyde and other known carcinogens.

Hill’s Postulates A list of epidemiological criteria that set the minimal evidence necessary to show causality.

Interventional study A study where one group is given a treatment to evaluate the effect of that treatment on people.

Meaningful relationship A type of relationship that cannot be shown to be causal, but has attempted to remove as many hidden variables as possible as an attempt to show causality.

Over-reported When survey respondents claim they partook of an activity more than they actually did.

Under-reported When survey respondents claim they partook of an activity less than they actually did.

Underlying variables Also known as “hidden” variables, are factors that obscure a relationship that is believed to be causal.