Related Experiment Videos
Statistical issues in randomized trials of cancer screening
Stuart G Baker1, Barnett S Kramer, Philip C Prorok
1Division of Cancer Prevention, National Cancer Institute, Bethesda, MD, USA. sb16i@nih.gov
BMC Medical Research Methodology
|September 20, 2002
Summary
This study introduces simple statistical methods for cancer screening trials. Recommendations include using cancer deaths as endpoints and adjusting for nonattendance and follow-up dilution.
Area of Science:
- Biostatistics
- Epidemiology
- Clinical Trials
Background:
- Randomized trials for cancer screening require unique statistical approaches distinct from therapeutic trials.
- Recent and novel methodologies are presented to address these specific challenges.
Framework:
- Focus on straightforward statistical methods for practical application.
- Recommendations are designed for clarity and ease of implementation in trial design and analysis.
Implementation:
- Utilize cancer-specific mortality as the primary endpoint, ensuring comprehensive reporting of all causes of death.
- Employ a "causal" estimation method to adjust for immediate post-randomization issues like nonattendance and contamination.
- Incorporate an adaptive estimation technique to manage dilution effects during extended follow-up periods after the final screening.
Implications:
- The proposed guidelines integrate current research on screening endpoints and noncompliance with a novel adaptive method for dilution adjustment.
- Adherence to these guidelines promotes robust practices in the design and statistical analysis of cancer screening randomized trials.