The strain-life designed is based on the damaged caused by the strain as a result of a loading history. The purpose of this approach is estimating the number of cycles to initiate a fatigue crack. Compared with the stress life method the EN needs a lower number of variables. However, it needs a large number of parameters that are not directly measurable properties of a material and conduct the necessary number of tests to obtain them is not always an option. This fact leads to the formulation of different models to estimate Coffin- Manson s constants, which in most of the cases can result in a non-conservative fatigue life prediction, as seen in several papers. However, in these studies, it is not considered a specimen with a stress concentration factor neither under a variable amplitude loading history. The methodology used in this research is evaluating the influence of these aspects by comparing the fatigue life experimental with the estimated by using the Coffin Manson s experimental constants and the fatigue life calculated with different models to estimate the Coffin-Manson s parameters. This study evaluated the fatigue damage based on four load histories and two types of samples with different stress risers. Considering the models of Manson, Muralidharan-Manson, Bäumel-Seeger, and Meggiolaro-Castro the results indicated that the fatigue life predicted using the Coffin Manson s parameter obtained by Meggiolaro-Castro and Baumel-Segger s model are more similar to the life calculated with the experimental constants. The other models are more influenced by the stress concentration s rule and can lead to not conservative fatigue life.