Statistical aspects of the design and testing of decompression tables.

Before a decompression procedure is recommended for general use it is subjected to a limited number of human trial dives. Based on the trial, one attempts to reject unsafe procedures but accept those with a low incidence of decompression sickness (DCS). Binomial confidence regions are often so broad that even after 40 dives it may be impossible to distinguish between the possibility that the table being tested has a 0.6% risk of DCS and the possibility that it has a 17% risk. Our proposed alternative is to select some rule (e.g., one or more cases of DCS in 10 dives) for rejecting tables and to calculate the probabilities of accepting tables as a function of the probability of DCS. With such calculations we conclude that (a) generally one cannot reduce the risk of adopting unsafe tables without increasing the risk of rejecting safe ones unless one chooses to increase the number of test dives; (b) truncated sequential designs could reduce the number of dives required for testing by 15 to 20%; and (c) rules similar to the ones tested will always have a zone of indifference. Tables with a probability of DCS in this zone will be accepted or rejected with nearly equal frequency even if tested with hundreds of dives. The use of models describing the probability of DCS as a function of dive parameters should allow us to combine information from dives previously analyzed separately and perhaps to improve our selection of new tables to be tested.