A physician at a university health clinic is designing an observational study to examine if regular consumption of energy drinks increases the risk of new-onset hypertension in college students. A major concern is that students who consume energy drinks may also have poorer sleep habits, and poor sleep is independently associated with higher blood pressure. To properly address this specific issue, which of the following methodological approaches should be considered? (Select all that apply)
- Limiting study enrollment strictly to students who report getting at least seven hours of sleep per nightAnswer
- Including sleep duration as a variable in a multivariable regression model during data evaluationAnswer
- CHiding the students' energy drink consumption status from the clinicians measuring their blood pressure
- DEnsuring that the enrolled students remain unaware of the specific hypothesis being tested
- EExpanding the target sample size to include a larger number of students across multiple university campuses
Answer
The correct approaches are limiting study enrollment to students with adequate sleep (restriction) and including sleep duration in a multivariable regression model.
Confounding occurs when a third variable is related to both the exposure and the outcome, distorting the true association. To control for confounding, researchers can use design-phase methods such as restriction (limiting the population to a single category of the confounder, thereby removing its variability) or matching. They can also use analysis-phase methods such as stratification or multivariable regression modeling (mathematically adjusting for the confounder). Both limiting enrollment based on sleep hours and utilizing regression modeling are valid techniques to isolate the independent effect of energy drinks on blood pressure.
Step-by-Step Solution
Key Concept
Methods to control for confounding variables in observational studies include design-stage strategies (matching, restriction) and analysis-stage strategies (stratification, multivariable regression).