Five-year change in statistical designs of phase II trials published in leading cancer journals