Clinical Trials Part 2

Question 1

What are the different types of outcome measures?

Answer 1

• categorical (qualitative)
• continuous (quantitative, numerical)

Question 2

What is the clinical effect size for continuous variable outcomes?

Answer 2

• clinical effect size is to evaluate w/ continuous outcomes
• authors of the paper will usually present the Means or Medians of the continuous outcomes in the various treatment grps
• ideally the 95% confidence interval of those means would also be presented to enable you to decide how much uncertainty should be included w/ their estimate of the mean (often they may only present the “standard deviation” or the “standard error” as a measurement of the variability rather than the confidence interval)
• difference btwn those means or medians is the “clinical effect size” & you need to subjectively consider that to determine if it is important or “clinically significant”

Question 3

How do you measure categorical variables and what are two issues with this?

Answer 3

Many clinical trials that deal w/ disease use an outcome measure that is CATEGORICAL
- mortality: dead or alive
- morbidity: sick or not sick
we now need to explore 2 issues:
1. how to measure the frequency of the health outcome event in the trial
2. how to measure the magnitude of the effect (impact) of the treatment or interventions

Question 4

How do you measure the magnitude of effect in clinical trials?

Answer 4

• we now have a way of counting the frequency of the outcome event (incidence or risk of disease during the trial)
• we now want to compare grps:
  > vaccinated & unvaccinated
  > treatment vs controls
• calculate the incidence or risk in both grps & compare risks

Question 5

What is the relative risk?

Answer 5

• index of the STRENGTH OF THE ASSOCIATION btwn the exposure & the disease (observational study)
• index of the MAGNITUDE OF THE CLINICAL EFFECT of treatment (clinical trial)
• RR = risk in controls / risk in treated
• by convention, ratio is set up so that the RR > 1.0, when possible, b/c it is easier to interpret

Question 6

What does RR = 1 mean?

Answer 6

means there is no treatment effect

Question 7

What does RR > 1 mean?

Answer 7

means the controls are at a greater risk of disease & the treatment group is protected by the treatment

Question 8

What does RR < 1 mean?

Answer 8

means the controls are at a reduced risk of disease & the treatment may actually make the disease worse

Question 9

What is attributable fraction?

Answer 9

• useful measure of clinical significance or in this case the vaccine efficacy
• what proportion of disease in the untreated animals could have been prevented by the vaccine / treatment?
• AF = (RR-1)/RR = (2-1)/2 = 50%
• therefore 50% of the mortality in the control grp could have been prevented by vaccination

Question 10

What is absolute risk reduction?

Answer 10

• amt in which your therapy reduces the risk of a bad outcome
• randomized controlled trial for strangles vaccination in Eq
• 20% of Eq receiving placebo developed strangles
• 12% of Eq receiving vx developed strangles
• RR: 20/12 = 1.67 (or 12/20 = 0.6)
• ARR = 20%-12% = 8%
• therefore if 100 Eq were vaccinated, 8 would be prevented from developing strangles

Question 11

what is the number needed to treat?

Answer 11

• different way of expressing absolute risk reduction or risk difference
• NNT = 1/ARR
• strangles vx ex:
  > NNT = 1/0.08 = 13 (always rounded up to the nearest whole number)
  > how many patients (Eq) do you need to treat (in this case vaccinate) to prevent one additional bad outcome?

Question 12

what is the odds ratio?

Answer 12

• estimate of the relative risk
• tends to over estimate the relative risk
• not commonly seen in clinical trials
• more commonly used in observational studies or field investigations (case-control studies)
• interpret in a similar fashion to relative risk
• odds of exposure in the diseased grp divided by the odds of exposure in the control grp (ad/bc)

Question 13

What is the role of chance using a statistical test?

Answer 13

• can now calculate the magnitude of the clinical effect for a 2x2 table (relative risk)
• need to be able to apply a statistical test to this to decide if this difference is due to chance
• chi-squared test (estimate of the “Fisher Exact Test”)
• as usual the starting point is the null hypothesis
• H”0”: there is no difference in the proportions btwn treatment & control grps

Question 14

What would the 2x2 table look like if there was no effect of vaccination on mortality?

Answer 14

• we would expect the same proportions of mortality (or risk) in the vaccinated versus the control grp
• if the overall mortality is 6 animals & we have exactly 100 animals in the controls & in the vaccinates, we would expect 2 to die in each grp if there was no effect of the vx