Estimating sample size for a randomized clinical trial of lung cancer screening.

BACKGROUND

Lung cancer screening, like other screening tests, has the potential to save many lives, yet its benefit has not yet been shown in randomized clinical trials (RCTs). RCTs of screening tests require very large sample sizes and long follow-up periods. Investigators planning these trials need to know how the study design, particularly the length of accrual, the number of incident screens, the length of follow-up, and the method used to interpret the screening test, affect sample size requirements.

METHODS

In this paper a tiered approach to sample size estimation is presented. The first tier determines the person-years at risk; the second tier considers the characteristics of the study population to estimate the frequency of preclinical disease; the third tier uses the performance of the screening test, the compliance rate, and the benefits of early treatment to estimate the incidence of clinical disease; the disease-specific mortality rates are estimated in the fourth tier.

RESULTS

This sample size approach was utilized in planning a RCT of lung cancer screening with chest X-rays and a computer aided detection algorithm. Sample size was determined for various features of the study design, as well as for different assumptions about the cure rate, the treatment complication rate, and the rate of potential misclassification of the primary endpoints.

CONCLUSION

The tiered approach to sample size estimation for RCTs of screening tests allows investigators to assess how features of the study design affect sample size requirements.