Virologic studies of human immunodeficiency virus type 1-infected patients have investigated either the emergence of resistance mutations according to the treatment received (type I) or their effect on subsequent regimens (type II). Type I studies provide an estimation of the frequency distribution of mutations for a given duration of therapy, but the delay to emergence of these mutations cannot be assessed. We suggest using a nonparametric estimator that generalizes the Kaplan-Meier method to data from type II studies to estimate the time to occurrence of mutations. Patients had no treatment interruption before viral genotyping. Although the curves should be interpreted with caution, they provide useful information about the kinetics of the emergence of mutations. The method was applied to the emergence of thymidine analogue mutations in patients previously treated with zidovudine (ZDV) plus didanosine or zalcitabine. Although K70R has been described as the first mutation to appear in patients receiving ZDV monotherapy, the T215Y/F mutation appeared first in patients receiving dual-nucleoside combination therapy.