[Skip to Content]
[Skip to Content Landing]
Views 339
Citations 0
Comment & Response
August 1, 2019

Design of Noninferiority Trials for Hypofractionated vs Conventional Radiotherapy Among Patients With Cancer

Author Affiliations
  • 1Department of Statistics and Actuarial Science, The University of Hong Kong, Hong Kong
  • 2Department of Biostatistics, MD Anderson Cancer Center, Houston, Texas
  • 3Department of Biostatistics, Harvard T. H. Chan School of Public Health, Boston, Massachusetts
JAMA Oncol. Published online August 1, 2019. doi:10.1001/jamaoncol.2019.2391

To the Editor Hypofractionated radiotherapy (HRT) is becoming increasingly popular in cancer treatment. Compared with conventional radiotherapy (CRT), HRT is delivered across shorter time frames in higher doses, potentially reducing costs and adverse effects while improving quality of life. Bruner et al1 demonstrated noninferiority of HRT to CRT among patients with low-risk prostate cancer in terms of disease-free survival. Wang et al2 conducted a similar noninferiority trial among patients with high-risk breast cancer. In both studies, the primary analysis was based on the difference in the 5-year survival rates; the study by Bruner et al1 used disease-free survival whereas the study by Wang et al2 used cumulative locoregional recurrence. The prostate cancer trial prespecified a noninferiority margin of 7%.1 Because the 5-year incidence of disease events with CRT was 14.7%, this design would have tolerated a 5-year incidence of 21.7% with HRT. Given the large noninferiority margin, the conclusion is unsurprising even before learning that the observed 5-year incidence of disease events was 13.7% in the HRT arm. Similarly, in the breast cancer trial, the 5-year locoregional recurrence with CRT was projected to be 6%, and the prespecified noninferiority margin was 5%.2 Thus, this design would have tolerated up to 11% locoregional recurrence in the HRT arm or almost double that of CRT. Given such a large noninferiority margin, the highly significant P value (P < .0001) is not surprising. The P value becomes increasingly significant as the noninferiority margin is increased. These examples highlight the importance of setting the noninferiority margin sufficiently narrow for the conclusion of noninferiority to be clinically meaningful.