Evolution of insect resistance threatens the continued success of transgenic crops producing Bacillus thuringiensis (Bt) toxins that kill pests. The approach used most widely to delay insect resistance to Bt crops is the refuge strategy, which requires refuges of host plants without Bt toxins near Bt crops to promote survival of susceptible pests. However, large-scale tests of the refuge strategy have been problematic. Analysis of more than a decade of global monitoring data reveals that the frequency of resistance alleles has increased substantially in some field populations of Helicoverpa zea, but not in five other major pests in Australia, China, Spain and the United States. The resistance of H. zea to Bt toxin Cry1Ac in transgenic cotton has not caused widespread crop failures, in part because other tactics augment control of this pest. The field outcomes documented with monitoring data are consistent with the theory underlying the refuge strategy, suggesting that refuges have helped to delay resistance.