A large randomised controlled trial is carried out to determine the effect of pravastatin, a cholesterol lowering medication, in elderly people (>65 years old) at risk of cardiovascular disease, i.e. heart attacks and strokes. Out of 23 770 individuals assessed at various family practices, only 7056 were eligible and agreed to participate. These were entered into a “run-in” period, in which they were given placebo tablets for 4 weeks. Those who were not compliant in taking their medicine were dropped from the trial, leaving 5804 patients to be randomised.
2913 were assigned placebo, and 2891 were assigned pravastatin. Over the next 3 years, 650 and 725 people discontinued their medication, respectively (lack of compliance), but all participants continued to be followed for fatal and non-fatal heart attacks, as well as fatal and non-fatal strokes. The authors perform an intention to treat analysis, with all cardiovascular events being the primary endpoint (combined fatal and non-fatal heart attacks and fatal and non-fatal strokes).
In the placebo group, 473 reached a primary endpoints vs. 408 in the pravastatin group.
1. Draw the 2 x 2 table for all cardiovascular outcomes combined (fatal and non-fatal heart attack and fatal and non-fatal stroke all combined).
2. What effect does the differential lack of compliance have? What is the direction of this bias? How does this affect internal validity?
3. What is the absolute risk of all cardiovascular outcomes in the placebo and drug group? Explain this in words?