From Association to Causation: Deriving Inferences from Epidemiologic Studies

Introduction to Causation and Epidemiology

• Understanding causal relationships in epidemiology is vital for:
  • Clinical decision-making regarding treatments and interventions.
  • Limiting exposure to hazardous agents (e.g., lead in paint).
  • Informing public health policies (e.g., laws on alcohol consumption, seat-belt use).

Types of Causal Relationships

• Risk Indicators vs. Causal Risk Factors:
  • Not all risk indicators are causal (e.g., age is a risk factor for cancer but does not cause it).
  • Causal risk indicators can be considered risk factors that contribute to disease.
  • Diseases result from combinations of genetic and environmental factors.

Causal Pathways: Direct vs. Indirect Effects

• Direct Effects:
  • A cause directly leads to a disease without intermediaries (e.g., lead exposure from cosmetics).
• Indirect Effects:
  • Factors like poverty can create conditions (e.g., substandard housing) that increase exposure to hazards (e.g., lead pipes), leading to disease.
• Web of Causation:
  • Shows various pathways and factors contributing to disease outcomes (e.g., lead exposure from multiple sources).

Causal Relationships Classification

• Necessary and Sufficient: Required for disease occurrence; e.g., having two defective alleles causes Tay-Sachs Disease.
• Necessary but not Sufficient: Some factors must be present for disease development but are not enough on their own (e.g., exposure to Mycobacterium tuberculosis for tuberculosis).
• Sufficient but not Necessary: Other factors also can cause the disease (e.g., genetic disorders like Tay-Sachs).
• Neither Necessary nor Sufficient: E.g., smoking is a risk factor but not everyone who smokes develops lung cancer.

Koch’s Postulates

• Criteria to identify causative agents of diseases:
  1. Organism always found in diseased individuals.
  2. Organism not present in healthy individuals.
  3. Isolated organism from infected individuals can cause disease in a host.
• Limitations include failing to account for carriers and inadequacies when applied to chronic diseases.

Multifactored Nature of Diseases

• Chronic diseases (e.g., cancer) stem from a mix of behavioral, environmental, and genetic factors.
• Rothman's model illustrates that each disease case can be a result of various risk factors.

Guidelines for Assessing Causality

• A framework proposed by Hill (1965):
  1. Temporal relationship: Exposure must precede the outcome.
  2. Strength of association: Stronger associations are more likely to be causal.
  3. Dose-response relationship: Increased exposure correlates with increased effect.
  4. Replication of findings: Consistent findings across studies strengthen causal claims.
  5. Biological plausibility: Mechanism should align with existing biological knowledge.
  6. Consideration of alternative explanations: Explore and rule out confounding factors.
  7. Cessation of exposure: Disease incidence should decline when exposure ceases.
  8. Consistency with existing knowledge: Findings should fit within broader scientific understanding.
  9. Specificity of association: Ideally, an exposure should correspond to a single outcome.

Example: Helicobacter Pylori and Duodenal Ulcers

• H. pylori has been shown to cause garstrings; evidence assessed through the causal guidelines.
  • Temporal: H. pylori correlates with chronic gastritis leading to ulcers.
  • Strength: Found in >90% of ulcer cases.
  • Dose-response: Higher density of bacteria aligns with ulcer occurrence.
  • Biological Mechanism: H. pylori damages mucosa, leading to acid susceptibility.

Models of Causality

• Evolution of models, emphasizing the multifactorial nature of diseases:
  1. Koch's postulates.
  2. Epidemiological triad (Host, Agent, Environment).
  3. Web of Causation.
  4. Rothman's Sufficient Cause Model.

Conclusion

• Understanding causal relationships enhances medical practice and public health policy.
• Causation is often complex, necessitating comprehensive assessment frameworks and consideration of multifactorial influences.