Evidence Based Medicine and the Role of Chance

Question 1

Explain what is meant by hypothesis testing in stats

Answer 1

• Used to answer a research questions
  
  • E.g. “Does additionally treating patients admitted to hospital with suspected MI with clopidogrel improve long term outcomes?”
• Scenario created that you don’t necessarily believe and you try to disprove it (Hypothetico-deductive method)
  
  • Null hypothesis (H0) = No effect of clopidogrel on outcomes
• If enough evidence is gathered to reject H0, we have grounds to favour our alternative hypothesis (often the research question)
  
  • Alternative (H1) = Clopidogrel improves outcomes (1-sided)
  • Alternative (H1) = Clopidogrel has an effect on outcomes (2-sided)

Question 2

What is the p value?

Answer 2

The probability, given that the null hypothesis is true, of obtaining data as extreme or more extreme than that observed

It is the result of a statistical test e.g. chi-square, t-test

The lower the p-value, the greater the statistical significance of the observed difference, meaning you reject the null hypothesis

A small p-value (typically less than 0.05) means there’s a low probability of observing the data if the null hypothesis is true

If the observed data is unlikely to have occurred under the null hypothesis, it suggests that the null hypothesis might be incorrect. This leads to rejecting the null hypothesis in favor of the alternative hypothesis, which proposes a different explanation for the observed results.

Question 3

Describe type 1 and type 2 errors

Answer 3

Type 1 error - Both null hypothesis true and reject the null hypothesis are concluded

Type 2 - Both null hypothesis false and fail to reject null hypothesis are concluded

Question 4

How can P-values be interpreted?

Answer 4

• 95% CI contains 1 - not enough evidence to reject H0, if doesn’t contain 1, evidence to reject H0
• 95% CI contain 0 - not enough evidence to reject H0, if doesn’t contain 0, evidence to reject H0

If p≥ 0.05 - Not enough evidence to reject H0

If P < 0.05 - Statistically significant

Question 5

Describe independent vs. paired data

Answer 5

Independent - 2 independent groups, interested in b/w group differences, no worry about variation b/w individuals in diff groups

Paired - Measured on same individuals, can provide more precise estimates of treatment effects

Question 6

Describe statistical correlation

Answer 6

• Statistical measure of (linear) relationship b/w 2 variables
  
  • Lies b/w -1 to 1
  • 0 implies no correlation
• Paired data often shows strong positive correlation

Question 7

What’s the difference between parallel and crossover clinical trials?

Answer 7

Parallel - One group receives treatment, other is control (receive placebo) and then the outcomes are measured

Crossover - Both groups will receive the treatment and placebo, and then the outcomes are measured. This focuses on short-term outcomes that are measurable in each period. Outcomes can then be compared within individuals

Question 8

Describe why crossover trials may be used

Answer 8

• Only suitable when treatment effects are short-lived and reversible
  
  • Requires washout periods b/w treatment periods help prevent ‘carry-over’ effects
  • Usually longer in duration vs. parallel
• Has some advantages over parallel
  
  • Smaller number subjects needed
  • Removes biological and methodological variations (as subjects act as their own control)
• Can be more than 2 treatments + 1 ‘crossover’