This paper presents an approach for performing a long-term fatigue analysis of rolling element bearings in wind turbine gearboxes. Multilevel integrated analyses were performed using the aeroservoelastic code HAWC2, the multibody dynamics code SIMPACK, the three-dimensional finite element code Calyx and a simplified lifetime prediction model for rolling contact fatigue. The National Renewable Energy Laboratory's 750 kW wind turbine and its planetary bearing were studied. Design load cases, including normal production, parked and transient load cases, were considered. To obtain the internal bearing load distribution, an advanced approach combining a finite element/contact mechanics model and a response surface model were used. In addition, a traditional approach, the Harris model, was also applied for comparison. The long-term probability distribution of the bearing raceway contact pressure range was then obtained using Weibull and generalized Gamma distribution functions. Finally, we estimated the fatigue life of the bearing, discussed the differences of the methods used to obtain the bearing internal loads and analyzed the effects of the environmental conditions and load cases on the results. The Harris model may underestimate the inner raceway life by 55.7%, which can cause large load fluctuations along the raceways. The bearing fatigue life is very sensitive to the wind distribution and less affected by the transient and parked load cases. Copyright © 2014 John Wiley & Sons, Ltd.