Fleming, TR and Harrington, DP (1991). Counting Processes and Survival Analysis. Wiley.
This two-arm survival program assumes both piece-wise exponential survival curves and piece-wise constant accrual over unit intervals.
The formulas are based on the large sample distribution of logrank test statistic, with non-centrality parameter calculated using piecewise uniform accrual over time, no loss to follow-up, and piece-wise exponentially distributed death times. The current program uses an approximation based on the expected number at risk at the midpoints of each interval. Therefore, increasing the number of accrual and follow-up intervals improves the performance of the approximation. In general this program should be used for approximate calculations then followed with appropriate simulation studies for more accurate properties.
The default is to estimate the sample size, using a fixed hazard ratio for all intervals. The user is prompted for the input listed below including the width of each interval and the number of accrual intervals and number of follow-up intervals. Note, because a patient can be followed up to total number of accrual (# accrual) and follow-up (# follow) intervals the hazard rates need to be filled in for a total of # accrual + # follow intervals. All input must be in terms of the same time units.
The user is prompted for values to the following items. For items that have initial default values set, the values are given in parentheses.