I took some time to expand my May 30, 2005 weblog entry on accrual rates and developed a seminar which I will present to the Statistics journal club at KUMC today.The handout for this talk combines that weblog entry with a brief tutorial on quality control.
I received some valuable feedback. One comment was on the use of the log transformation, and I might want to show this graph on a different scale. Another comment was on the use of Bayesian statistics to model accrual rates. As the data accumulates, you would get more information about the estimated stopping date of the study or about the estimated total sample size at the end of the study. It would be important to get a good prior distribution or possibly both an optimistic and a pessimistic prior. If even an optimistic prior is swamped by actual data suggesting that accrual is slow enough that no realistic sample size would emerge at the of the study, then that might be the point at which you decide to cut your losses.
You can find an earlier version of this page on my old website.