Mechanical components are very commonly subjected to nontrivial, time-variant stress-states during their operations, and therefore must be designed adequately in order to prevent failures due to fatigue. Given the complex nature of non-uniaxial stress-states, it is important to consider a robust and adequate theoretical approach that will ensure the safety of engineering designs. The present study aims to assess the predictive capability of a new multiaxial fatigue criterion proposed by Castro, Pereira and Darwish, henceforth named Elliptical Method (E), wherein the fatigue lives of specimens subjected to synchronous sinusoidal axial and torsional loadings are evaluated. These specimens are manufactured from DIN 42CrMo4 steel. The assessment is carried out by comparing the predictive capability of the elliptical method (E) with the predictions obtained through other criteria available in the literature, namely Findley (F), Matake (M), McDiarmid (McD), Susmel & Lazzarin (S&L), Carpinteri & Spagnoli (C&S) and Papadopoulos (P). With the exception of the C&S criterion, which has been previously adapted by the authors themselves in order to predict fatigue-life, this study appropriately adapts the other criteria to predict the number of cycles to failure of a certain specimen under a given loading condition. Thus, all the predictions (including the ones obtained through the elliptical method) are compared to experimental observations relative to multiaxial tests carried out with the aforementioned specimens. The comparison reveals that the C&S criterion is the one that presents the most accurate prediction capability, while the elliptical method achieves second place. Yet, the C&S criterion shows some problems: it presents a slightly non-conservative behaviour since some of its fatigue-life predictions appear higher than those observed experimentally. Also, this particular criterion requires the most extensive procedure in order to obtain the number of cycles to failure. On the other hand, although not as precise as the C&S criterion, the elliptical method presents the second most accurate predictive capability, with the advantages of being approximately 24% conservative, as well as being the easiest criterion to apply among all the considered models. This criterion (E) also presents reasonably accurate results when compared to experimental observations.