Abstract A variety of design and patient parameters have been implicated in recent reports of fretting corrosion at modular connections in total hip arthroplasty. We sought to identify the relative sensitivity of mechanical fretting to a comprehensive set of parameters such that attention may be focused on key variables. Stochastic finite element simulation of the head–neck taper–trunnion junction was performed. Four-hundred parameter sets were simulated using realistic variations of design variables, material properties and loading parameters to predict contact pressures (P), micromotions (M) and fretting work (coefficient of friction×P×M) over cycles of gait. Results indicated that fretting work was correlated with only three parameters: angular mismatch, center offset and body weight (r=0.47, 0.53 and 0.43, p<0.001). Maximum contact pressure increased by 85MPa for every 0.1° of angular mismatch. Maximum micromotion increased by 5µm per 10mm additional head offset and 1µm per 10kg increased body weight. Uncorrelated parameters included trunnion diameter, trunnion length and impaction forces. It was concluded that appropriate limiting of angular mismatch and center offset could minimize fretting, and hence its contribution to corrosion, at modular connections.