Laboratory bench tests have shown that porosity reduction increases the fatigue life of bone cement specimens. Clinically, however, the effect porosity reduction is subject to debate. We hypothesized that the discrepancy between clinical and experimental findings is related to differences in the stress distribution, which is typically uniform in experimental test specimens, while stress concentrations exist in cement around hip implants. We simulated fatigue failure of cement in a finite element model of an experimental test specimen and of a transverse slice of a total hip arthroplasty with a sharp-cornered stem. Four levels of porosity were introduced. In the fatigue test specimen model, the fatigue life clearly was dependent on the level of porosity, while in the transverse slice model, the level of porosity had virtually no effect on failure of the cement mantle. The results of the simulations confirmed our hypothesis. In simulations of laboratory tests, pores clearly acted as crack initiators, while in the simulation of a real total hip reconstruction, crack formation was governed by local stress singularities. This explains why the beneficial effect of cement porosity reduction on the lifetime of total hip reconstructions may be hard to detect clinically.