In the context of emerging stacked intelligent metasurface (SIM)-based holographic MIMO (HMIMO) systems, a fundamental problem is to study the mutual information (MI) between transmitted and received signals to establish their capacity. However, direct optimization or analytical evaluation of the MI, particularly for discrete signaling, is often intractable. To address this challenge, we adopt the channel cutoff rate (CR) as an alternative optimization metric for the MI maximization. In this regard, we propose an alternating projected gradient method (APGM), which optimizes the CR of a SIM-based HMIMO system by adjusting signal precoding and the phase shifts across the transmit and receive SIMs on a layer-by-layer basis. Simulation results indicate that the proposed algorithm significantly enhances the CR, achieving substantial gains, compared to the case with random SIM phase shifts, that are proportional to those observed for the corresponding MI. This justifies the effectiveness of using the channel CR for the MI optimization. Moreover, we demonstrate that the integration of digital precoding, even on a modest scale, has a significant impact on the ultimate performance of SIM-aided systems.