Skip to main content
Figure 4 | Trials

Figure 4

From: Predicting clinical trial results based on announcements of interim analyses

Figure 4

Screenshot of Excel spreadsheet demonstrating the calculations described in the text. Entry ‘5’ at cell 10B refers to the entry in cell 7A, and sets the prior probabilities to those labeled ‘Custom’ in row 7. These were the probabilities of the author (DAB) during the early conduct of National Surgical Adjuvant Breast and Bowel Project trial C-08 (NSABP C-08), and were based on the publicly available information about bevacizumab and its potential efficacy in treating adjuvant colon cancer, as described in the text. ‘Interim analysis 7’ is actually the final analysis. The emboldened type in row 18 indicates the probabilities of eventual success, futility, and failure after the fourth interim analysis. This spreadsheet is available at http://www.berryconsultants.com/wp-content/uploads/2014/02/Unknown-Hazard-Reduction.xls. It allows users to input their own probabilities, and to experiment with other prior probabilities to see how they affect conclusions. The column headings are ‘suc’ (the probability of success at the current row’s interim analysis); ‘fut’ (the probability of futility at the current row’s interim analysis); ‘undec’ (the probability of making no decision at the current row’s interim analysis and continuing the study); ‘suc|undec (the probability the trial will be declared successful at that row’s interim analysis, given no decision at any previous analysis); ‘fut|undec (the probability the trial will be declared futile at that row’s interim analysis, given no decision at any previous interim analysis); ‘eventual suc|undec (the probability the trial will be declared successful at the current, future, or final analysis, given no decision at any previous analysis); ‘eventual fut|undec (the probability the trial will be declared futile at the current, future, or final analysis, given no decision at any previous analysis); and ‘eventual fail|undec (the probability the trial will fail to show statistical significance given no decision at any previous analysis).

Back to article page