Gordis Chapter 11
Chapter 11: Randomized Trials: Some Further Issues
Learning Objectives
Define key concepts of epidemiologic study design in randomized trials:
Sample size
Type I error
Type II error
Power
Generalizability (external validity)
Internal validity
Calculate and interpret efficacy in a randomized trial.
Describe the design and results of five historically important randomized trials.
Define the four major phases of randomized trials used by the US Food and Drug Administration for evaluating new drugs.
Introduce ethical considerations relating to randomized trials.
Discuss the rationale for registering new randomized trials in advance.
Sample Size
Importance of sample size:
Critical to determine before conducting a study.
Conducting a study with too few subjects can lead to meaningless results.
Example: Investigator's anecdote about a sheep study highlights the necessity of proper sample size to observe true differences.
Visual aid: Jars of beads analogy illustrates that small sample sizes could lead to incorrect conclusions about larger populations.
Types of Errors in Trials
Four Possible Study Outcomes:
Correct conclusion of no difference when there is none.
Incorrectly concluding a difference when there is none (Type I error).
Incorrectly concluding no difference when there is one (Type II error).
Correct conclusion of a difference when it exists.
Type I Error (α): The probability of concluding differences exist when they do not.
Type II Error (β): The probability of failing to conclude differences exist when they do.
Power (1-β): The probability of correctly concluding differences exist when they do.
Estimating Sample Size
Specify the expected difference in response rate.
Estimate the response rate in one group.
Determine level of statistical significance (α).
Choose desired power (1 - β).
Decide if the test should be one- or two-sided.
Recruitment and Retention
Challenges in recruitment: Must ensure enough eligible volunteers without coercing them.
Retention issues: Loss of participants over time can bias results. Investigators should understand reasons for dropouts and plan interventions to retain participants.
Ways to Express Results
Efficacy can be calculated as:
Efficacy = (Rate in placebo - Rate in treatment) / Rate in placebo.
Common terms:
Efficacy vs. Effectiveness
Relative risk comparisons
Number Needed to Treat (NNT) calculation.
Generalizability and Validity
External Validity: Ability to generalize trial results to a broader population.
Internal Validity: Ensures that outcomes truly reflect the treatment effect, free from biases.
Randomized trials generally have high internal validity but may suffer in external validity due to selection of healthier participants.
Registration of Clinical Trials
Registration helps combat publication bias by ensuring that all trials are documented.
The International Committee of Medical Journal Editors requires clinical trials to be registered before participants are enrolled.
Ethical Considerations
Discusses ethical dilemmas regarding randomization, informed consent, and the withholding of effective treatments.
Emphasizes that randomization is ethically sound only when there is uncertainty about treatment efficacy.
Conclusion
Randomized trials remain essential for evaluating medical therapies; they need rigorous design to ensure valid generalizable results.