What would be the analog of the "rule of three", but for a 99% confidence interval?

I.e., how to estimate, with the 99% confidence, the rate of occurrences in the population after looking at a sample with n subjects which didn't have the given event occurring?