Lecture 8 - Causation + Screening

How do you prove causation in non-infectious diseases?

Drafted by the US Surgeon General’s committee to address the possible relationship between smoking and lung cancer:

• Temporal relationship

• Strength of the association

• Dose-response relationship

• Replication of the findings

• Biological plausibility

• Consideration of alternate explanations

• Cessation of exposure

• Consistency with other knowledge

• Specificity of the association

What is a temporal relationship?

• Comes before the actual effect

• Establish sequence of events (A → B or B → A)

• Establish the interval between exposure and disease

  • Can be multifactorial

• Easiest to establish in prospective cohort studies

• Ex. Smoke levels in the air compared to the daily deaths follow a similar pattern

What is the strength of association?

• Measured by the relative risk

• Higher relative risk, higher likelihood of risk

• Not absolute - confounding factors

What is the dose-response relationship?

• How the response changes in proportion to the varying dose

• Not optimal because it does not necessarily rule out causation if not present

  • Diseases where there may be a binary threshold

• Ex. how many packs smoked a day versus mortality rate

What is the replication of the findings?

• Multiple studies that have the same conclusion make it more likely

• Although all may have confounding factors

• Correlation /= causation

What is biological plausibility?

• Current scientific knowledge may not be sufficient

• Plausible mechanism?

• Ex. Per capita cheese consumption vs number of people who died by becoming tangled in their bedsheets correlates

What is the consideration of alternate explanations?

• Not one thing speaks entirely, never flawless

• Totality of evidence

• Scientists design an experiment to disprove a hypothesis

What is cessation of exposure?

• Completely stopping contact with a harmful substance or agent, like a chemical, allergen, or tobacco smoke, to prevent further harm or disease

• Ex. Acute muscle pain and the FDA recalling tryptophan

What is consistency with other knowledge?

• information must cohere and not fundamentally contradict established, reliable knowledge or evidence

• Related to biological plausibility and temporal relationship

• Ex. Trends between cigarette consumption and lung cancer in men and women

What is an example of the benefits of the committee’s guidelines?

• Gastric ulcers

• It was consistent with other studies that gastric ulcers were caused by stress, coffee, etc.

• Dr. Warren found bacteria in ulcer biopsy samples, Helicobacter pylori

• No one believed him because they already believed that it was caused by what was previously known, especially since changes in diet and lifestyle decreased ulcers

  • How could bacteria survive in the acidic stomach

• Alternative explanations for H. pylori

  • The biopsy specimens were contaminated AFTER the samples were taken from the patients

  • The bacteria live in the stomach, but do no damage

  • The bacteria are an opportunistic species that arrives AFTER ulcers have already weakened the stomach’s defences

• To prove cause and effect, Dr. Marshall swallowed a flask of H. pylori from the lab

  • Within a week, he was suffering from symptoms of gastritis, and he had H. pylori in his stomach

  • He cured himself with an antibiotic treatment

How does H. pylori survive in the stomach?

• Converts urea to ammonia, raising pH (decreasing acidity)

• Flagella allow it to move to the stomach lining (neutral pH)

• Secretes toxin VacA that damages the stomach lining

• H. pylori grows in the mucous layer → inflammation → ulcers

What modifications to the committee’s guidelines were taken for causal inferences?

• Prenatal influences on child development can be hard to establish causality

• In 1989, the US Public Health Service modified the guidelines to include:

  • Categorizing the evidence by the quality of its source

    • Evolved from HRT

    • Randomized blinded trials → case studies

  • Prioritizing criteria for causality

    • Temporal relationship

    • Biological plausibility

    • Consistency

    • Alternate explanations

What was the US Preventive Services Task Force?

• A grading system for the suggestion of practice based on the magnitude of net benefit

• Grade A & B → Offer/provide service

• Grade C → Offer/provide service only if there are other considerations in support of the service

• Grade D → Discourage the use of the service

• Grade I → If offered, patients should understand the uncertainty about the balance and benefits of harm

What is GRADE?

• Grading of Recommendations Assessment, Development and Evaluation

  • Used by the WHO and more countries than the US

• Quality of evidence and definitions

  • High → Very confident that the true effect lies close to the estimate of the effect

  • Moderate → We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different

  • Low → Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect

  • Very Low → We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimated effect

What is disease progression and screening?

• Early detection does not guarantee any change in outcome

• Some cancer diagnoses, no matter when found, cannot be helped

What is an example that exhibits disease progression and screening?

• Breast cancer

• It appears that the lower mortality rate is due to the introduction of breast cancer screening, but it is actually due to better treatment

What is an issue with guidelines?

• National ones may not match provincial ones

• Ex. For breast cancer, Ontario says to start getting mammograms at 40, while Canada says mammograms at age 40-49; the harms may outweigh the benefits

  • With exceptions like family history, breast density, and ethnicity

What are the harms of screening?

• Analogy: Deer-vehicle collisions in NE USA result in >1 million collisions per year with >200 deaths

• What if you could prevent 21,400 injuries and 155 deaths by an intervention? Would you recommend it?

• What if you could prevent 21,400 injuries and 155 deaths by an intervention, but the intervention would directly harm 30 people? Would you recommend it?

• Would it be logical/ethical to debate the value of the intervention without mentioning the 30 people harmed?

• For screening, harms need to be discussed since they are typically hidden/unknown to the public

What are the potential outcomes of screening?

• Lead Time Bias

  • That the earlier diagnosis creates a better survival rate when, in reality, the usual time of diagnosis and treatment would result in the same time of death

• Overdiagnosis Bias

  • Due to a larger population being screened, people were overdiagnosed, which caused the survival rate to appear higher, but there was no survival rate change for those who had the disease

How does the validity of screening tests work?

• Many tests are dichotomous (binary) - result is either positive or negative

• Validity - for regulatory approval, can test distinguish disease and control

  • Needs to determine how accurate and how well they perform in a clinical setting

• Sensitivity - the ability of the test to identify correctly those who have the disease

  • Need to know if the patient has the disease or doesn’t

• Specificity - ability of the test to identify correctly those who do not have the disease

What is the comparison of the results of a dichotomous test with disease status?

• True Positive → Have the disease and test positive

• False Negative → Have the disease but test negative

• False Positive → Do not have the disease but test positive

• True Negative → Do not have the disease and test negative

How do binary tests work for continuous variables?

• By comparing the means or distributions of the continuous outcome between the two groups defined by the binary variable

• Ex. Blood sugar levels in persons with diabetes

  • The cutoff level is problematic because some patients with diabetes fall under the cutoff

What is the predictive value of a test?

• The probability that the test results are positive in a patient that they have the disease

• Most useful metric for diagnostic tests