Responsible of Statistical Analysis
Dr. Valter Torri
Responsible of Laboratory of Methodology for Clinical Research
Institute of Pharmacological Research IRCCS Mario Negri
Milan
Statistical Analysis
The primary endpoint is the pCR rate. Based on the results published by prospective studies on the delayed or observational time interval and retrospectives of the standard time interval, it has been assumed that the average percentage of pCR in standard treatment is on average 15, while in the longer time is 30. To determine this difference, 270 patients are needed, using the two-group chi-panel test with continuity correction, for the study of proportions, assuming an alpha error of 4.9 and a power of 80 (MedCalc version 17.9.7); an interim effectiveness analysis will be performed when half of the events are observed. The Haybittle-Peto method will be used as a break criterion in order to perform the final analysis with a final significance of 4.9% Considering that the percentage of unfavorable patients was calculated to be 20 (the degree of favorable MRI tumor regression is defined as grades 1, 2 and 3; the unfavorable MRI regression as grades 4 and 5), and that in calculating the sample size, it has been assumed that the percentage of missing data will be 5, a total of 340 patients has been expected, 170 for each arm. Patients will be randomized on a 1:1 basis. Block randomization software will be used.
Reference: Haybittle JL repeated assessment of results in clinical trials of cancer treatment. Be j radiol 1971, 793-797
