Extrapolation of IUD continuation curves

Summary Calculation of the number of contraceptive acceptors needed to register predesignated effects on birth or growth rates presupposes that it is known for how long couples practise each contraceptive method. Life table techniques may be used to estimate proportions continuing with a given method: but most follow-up studies yield data for durations no longer than from two to three years. To estimate continuation rates for the longer durations relevant to a five-or ten-year target period, the curve of continuation must be extrapolated. Four functions, including the one most commonly used and three new ones, are compared on the basis of data from the Taichung IUD Follow-up Study that commands an effective observation period of eight years. This unusual length makes it possible to perform the experiment of pretending that observation length is only one, two, … five years and then compare results based on these varyingly abbreviated spans with those based on the full observation period. No one of the four functions is found to be ideal under all circumstances. A strategy for choosing among the functions is suggested.