Effective stress measures are utilized by engineers to correlate uniaxial and biaxial experimental rupture data to creep failure in real components. Different effective stress formulations are used in various high temperature design codes, such as the Section III Division 5 of the ASME Boiler and Pressure Vessel code and the French RCC-MRx, and in different fitness for service provisions, such as the British R5 and the American FFS-1. Under uniaxial and biaxial stress states all the investigate metrics predict similar rupture times and provide similar safety margins, however there is no guarantee that when these metrics are extrapolated to predict the creep life for triaxial state they will remain conservative. In this work we use the real geometry of a Grade 91 end-plate that failed prematurely and finite element simulations to investigate the variability in creep life predictions obtained utilizing different effective stress as creep-damage driving force. Results show that for triaxial stress states the time to crack nucleation, the crack initiation site and the crack path are extremely sensitive to the effective stress measure.