Chapter 14 Epidemiology

Final Exam review sheet 

Ch 14    From Association to Causation: Deriving Inferences from Epidemiologic Studies 

Define, explain, and recognize the following in examples: 

Spurious association: A false relationship between two variables caused by a third factor. 
Causality: The direct cause-and-effect relationship between two variables. 
In vitro study: Experiments conducted in a controlled environment outside a living organism, like in a lab. 
“Unplanned” experiments: Experiments that occur without prior design or control, often in natural settings. 

Disease etiology: The study of the causes or origins of diseases. 
Confounding: When a third factor distorts the true relationship between two variables. 
Bias: A systematic error that leads to incorrect conclusions. 
Web of causation: A model showing how multiple factors interact to cause a disease. 

Rothman’s Pie Chart – component causes: A diagram showing how different factors contribute to disease, with each factor being a component cause. 
Spurious association: A false or misleading relationship between two factors due to a third variable. 

Necessary factor in disease development: A factor that must be present for a disease to occur. 
Sufficient factor in disease development: A factor that can cause a disease on its own. 

Answer/explain the following: 

What are the advantages and disadvantages of using animal studies for determining disease etiology? 

Advantages of animal studies: 

• Controlled exposure and environment 

• Minimal loss to follow-up 

• Early testing before human exposure 

Disadvantages of animal studies: 

• Uncertain if results apply to humans 

• Some human diseases don’t occur in animals 

• Hard to compare animal and human doses 

• Species may respond differently 

What are the advantages and disadvantages of using in vitro studies for determining disease etiology? 

Advantages: 
Allows researchers to study disease processes in a controlled and focused way. These systems are useful for testing how cells react to certain substances without involving animals or humans. 

Disadvantages: 
Because in vitro systems are artificial, it's hard to know if the results apply to real human bodies. They don't show how a substance affects the whole human organism, so the findings may not fully represent how a disease works in people. 

When is it NOT appropriate to conduct human population randomized trials when determining disease etiology? 

Ethical Reasons 

1. Harmful Exposures: Can't deliberately expose people to suspected harmful agents (smoking, radiation) 

2. Withholding Benefits: Unethical to deny known effective treatments 

3. Vulnerable Groups: Special protections needed for children, pregnant women 

Practical Reasons 

4. Rare Diseases: Impractically large sample sizes needed 

5. Long Latency: Can't wait decades for disease development (cancer) 

6. Widespread Exposures: Impossible to create unexposed control groups 

7. Complex Diseases: Multiple interacting causes difficult to isolate 

Methodological Reasons 

8. Compliance Issues: Long-term behavioral changes difficult to maintain 

9. Past Exposures: Can't randomize historical events 

10. Genetic Factors: Genetic traits cannot be randomly assigned 

What is the difference between association and causality? 

Association means two things happen together. 
Causality means one thing actually causes the other. 

What tends to come first in the normal practice of epidemiology – association or causation?  Explain your answer. 

In normal practice, association tends to come first. This is because researchers typically start by observing a link between two factors (e.g., smoking and lung cancer) before they can determine if one actually causes the other. Establishing causation requires more evidence and testing, but finding an association is the first step in identifying potential relationships. 

What are the four types of causal relationships?  Explain each and give an example of each. 

1. Necessary and Sufficient 

1. Explanation: This is when the factor is both required and enough on its own to cause the disease. 

2. Example: HIV is necessary and sufficient to cause AIDS—you need HIV for AIDS to happen, and having HIV alone can lead to AIDS. 

2. Necessary, but Not Sufficient 

1. Explanation: The factor is required, but there are other factors that also contribute to the disease. 

2. Example: Smoking is necessary for lung cancer, but it alone doesn’t always cause cancer—other factors, like genetics or air pollution, are also involved. 

3. Sufficient, but Not Always Necessary 

1. Explanation: This factor alone can cause the disease, but it’s not the only possible cause. 

2. Example: Infection with the flu virus is sufficient to cause flu symptoms, but there are other ways people can get sick without this specific infection. 

4. Neither Sufficient Nor Necessary 

1. Explanation: This factor alone is not enough to cause the disease, and it’s also not required for the disease to occur. 

2. Example: Stress may contribute to heart disease, but it isn’t enough on its own to cause it, nor is it required—other factors like diet, exercise, and genetics also play a role. 

Necessary and Sufficient  

Factor A = Disease 

Necessary, but Not Sufficient 

Factor A + Factor B + Factor C = Disease 

Sufficient, but Not Always Necessary 

Factor A or Factor B or Factor C = Disease 

Neither Sufficient Nor Necessary 

Factor A or Factor B + Factor C or Factor B = Disease 

What are the guidelines for judging whether an observed association is causal?  Explain each and give an example of each.  Be able to recognize how study results may contribute to a criterion. 

1. Temporal Relationship 

Explanation: The cause must happen before the effect. Example: People must start smoking before developing lung cancer, not after. 

2. Strength of the Association 

Explanation: Stronger connections are more likely to be causal. Example: Smokers have a 20-30 times higher risk of lung cancer than non-smokers. 

3. Dose-Response Relationship 

Explanation: More exposure leads to stronger effects. Example: People who smoke more cigarettes per day have higher cancer risk. 

4. Replication of the Findings 

Explanation: Different studies by different researchers find the same results. Example: Multiple studies in different countries all show lead exposure damages brain development. 

5. Biologic Plausibility 

Explanation: The relationship makes sense based on what we know about biology. Example: UV radiation damages DNA, so it's biologically plausible that sunlight causes skin cancer. 

6. Consideration of Alternate Explanations 

Explanation: Other possible explanations have been ruled out. Example: Studies on coffee and heart disease must account for whether coffee drinkers also smoke more. 

7. Cessation of Exposure 

Explanation: When exposure stops, risk decreases. Example: When people quit smoking, their lung cancer risk gradually decreases. 

8. Consistency with Other Knowledge 

Explanation: The relationship fits with what we already know. Example: The link between bacteria and stomach ulcers fits with how infections work elsewhere in the body. 

9. Specificity of the Association 

Explanation: The cause is linked to a specific effect, not many random ones.  

Example: Asbestos exposure specifically causes mesothelioma, a rare cancer that is almost exclusively linked to asbestos exposure. This high specificity in both directions strengthens the causal relationship. 

Measuring the Strength of Association Between Exposure and Disease 

The strength of association between an exposure and disease is measured using several epidemiological measures: 

1. Relative Risk (RR): The ratio of the incidence rate in exposed individuals to the incidence rate in unexposed individuals. An RR of 1.0 means no association, while values greater than 1.0 suggest increased risk. 

2. Odds Ratio (OR): Compares the odds of exposure in those with disease to the odds of exposure in those without disease. Often used in case-control studies when incidence cannot be calculated. 

3. Attributable Risk (AR): The difference in disease rates between exposed and unexposed groups, showing how much disease can be attributed to the exposure. 

4. Population Attributable Risk (PAR): The proportion of disease in the population that would be eliminated if the exposure were removed. 

5. Hazard Ratio (HR): Similar to relative risk but accounts for time-to-event in survival analyses. 

U.S. Public Health Service Process for Causal Inference (1989) 

The basic process developed by the U.S. Public Health Service for causal inference includes: 

1. Evidence Gathering: Collecting all relevant studies and data related to the exposure-disease relationship. 

2. Quality Assessment: Evaluating the methodological quality of available studies. 

3. Consistency Evaluation: Determining whether findings are consistent across different studies and populations. 

4. Strength Assessment: Examining the magnitude of associations found in studies. 

5. Expert Panel Review: Having specialists review the evidence and form conclusions. 

6. Categorization of Evidence: Classifying evidence as sufficient, suggestive, inadequate, or suggestive of no association. 

7. Recommendation Development: Creating public health recommendations based on the strength of causal evidence. 

Hierarchy of Study Designs (Highest to Lowest Quality) 

1. Systematic Reviews and Meta-analyses  

1. Highest quality because they synthesize multiple studies, reducing bias and increasing precision. 

2. Randomized Controlled Trials (RCTs)  

1. High quality due to randomization, which minimizes selection bias and balances known and unknown confounders between groups. 

3. Cohort Studies  

1. Track exposed and unexposed groups forward in time, establishing temporal sequence but vulnerable to loss to follow-up and confounding. 

4. Case-Control Studies  

1. Compare disease cases with controls, efficient for rare diseases but subject to recall bias and selection bias. 

5. Cross-sectional Studies  

1. Measure exposure and outcome at the same time, cannot establish temporality, making causal inference difficult. 

6. Case Series and Case Reports  

1. Descriptive studies without comparison groups, useful for hypothesis generation but not for establishing causation. 

7. Expert Opinion  

1. Lowest quality because it relies on individual judgment rather than systematic research. 

This hierarchy is based on each design's ability to minimize bias and confounding factors, which strengthens causal inference. 

Role and Process of the U.S. Preventive Services Task Force (USPSTF) 

Role 

The USPSTF is an independent panel of experts that systematically reviews evidence and develops recommendations about preventive health services such as screenings, counseling, and preventive medications. 

Process 

1. Topic Selection: Identifies preventive services for review based on public health importance. 

2. Research Plan Development: Creates an analytic framework and key questions. 

3. Evidence Review: Commissions systematic evidence reviews from Evidence-based Practice Centers. 

4. Evidence Evaluation: Assesses the quality, strength, and limitations of available evidence. 

5. Recommendation Development: Weighs benefits against harms to determine net benefit. 

6. Draft Recommendation: Issues preliminary recommendations for public comment. 

7. Final Recommendation: Publishes final recommendations after considering feedback. 

USPSTF Grading System for Net Benefit 

The USPSTF presents the balance of net benefit using letter grades: 

Grade A 

• Meaning: High certainty that the net benefit is substantial. 

• Recommendation: Offer or provide this service. 

• Usage: Strong evidence for implementation in clinical practice. 

Grade B 

• Meaning: High or moderate certainty that the net benefit is moderate to substantial. 

• Recommendation: Offer or provide this service. 

• Usage: Generally recommended for appropriate patients. 

Grade C 

• Meaning: Moderate certainty that the net benefit is small. 

• Recommendation: Selectively offer based on professional judgment and patient preferences. 

• Usage: Consider for individual patients based on specific circumstances. 

Grade D 

• Meaning: Moderate or high certainty that there is no net benefit or that harms outweigh benefits. 

• Recommendation: Discourage the use of this service. 

• Usage: Avoid implementing in clinical practice. 

Grade I Statement 

• Meaning: Current evidence is insufficient to assess the balance of benefits and harms. 

• Recommendation: If offered, patients should understand the uncertainty about benefits and harms. 

• Usage: Clinical decision-making should incorporate other factors and patient preferences. 

These grades are used by: 

• Healthcare providers to guide clinical decisions 

• Healthcare organizations to develop practice guidelines 

• Insurers for coverage determinations 

• Policymakers for healthcare program planning 

• Patients and clinicians for shared decision-making 

The grading system ensures transparent communication about the level of confidence in preventive service recommendations and helps prioritize services with the greatest potential to improve public health.