Stopping rules employing response rates, time to progression, and early progressive disease for phase II oncology trials
Journal Articles
Overview
Research
Identity
Additional Document Info
View All
Overview
abstract
BACKGROUND: Response rate (RR), the most common early means of assessing oncology drugs, is not suitable as the sole endpoint for phase II trials of drugs which induce disease stability but not regression. Time to progression (TTP) may be more sensitive to such agents, but induces recruitment delays in multistage studies. Early progressive disease (EPD) is the earliest signal of time to progression, but is less intuitive to investigators, To study drugs with unknown anti-tumour effect, we designed the Combination Stopping Rule (CSR), which allows investigators to establish a hypothesis using RR and TTP, while the program also employs early progressive disease (EPD) to assess for drug inactivity during the first stage of study accrual. METHODS: A computer program was created to generate stopping rules based on specified error rates, trial size, and RR and median TTP of interest and disinterest for a two-stage phase II trial. Rules were generated for stage II such that the null hypothesis (H(nul)) was rejected if either RR or TTP met desired thresholds, and accepted if both did not. Assuming an exponential distribution for progression, EPD thresholds were determined based on specified TTP values. Stopping rules were generated for stage I such that Hnul was accepted and the study stopped if both RR and EPD were unacceptable. RESULTS: Patient thresholds were generated for RR, median TTP, and EPD which achieved specified error rates and which allowed early stopping based on RR and EPD. For smaller proportional differences between interesting and disinteresting values of RR or TTP, larger trials are required to maintain alpha error, and early stopping is more common with a larger first stage. CONCLUSION: Stopping rules are provided for phase II trials for drugs which have either a desirable RR or TTP. In addition, early stopping can be achieved using RR and EPD.
