There is a recent thesis in the literature that an important organization of a physical system precedes a catastrophic event. In this context, one can search for signatures that imply the transition from a normal state to a main catastrophic event (e.g., earthquake). Experimental techniques are thus useful in corroborating theories from observed data. For example, recent results indicate that preseismic electromagnetic time series contain information characteristic of an ensuing earthquake event. Hereby, we attempt to demonstrate that an easily computable complexity measure, such as T-complexity or approximate entropy, gives evidence of state changes leading to the point of global instability. The appearance of a precatastrophic state is characterized by significant lower complexity in terms of T-complexity and approximate entropy. The present study confirms the conclusions of previous works based on an independent linear fractal spectral analysis. This convergence between nonlinear and linear analysis provides a more reliable detection concerning the emergence of the last phase of the earthquake preparation process. More precisely, we claim that our results suggest an important principle: significant complexity decrease and accession of persistency in electromagnetic (EM) time series can be confirmed at the tail of the preseismic EM emission, which could be used as diagnostic tools for the Earth's impending crust failure. Direct laboratory and field experimental data as well as theoretical arguments support the conclusions of the present analysis.