This study aimed to optimize the collimator length for a parallel-hole prompt gamma camera dedicated to real-time proton beam range verification, achieving an optimal balance between detection sensitivity and spatial accuracy for clinical application. Comprehensive Monte Carlo simulations were conducted using the GATE platform with the QGSP_BIC_HP_EMZ physics list. A cylindrical PMMA phantom was irradiated with a 90 MeV monoenergetic proton beam, and a PG camera model consisting of an LSO scintillator coupled with a lead parallel-hole collimator was simulated. Collimator length was systematically varied from 4 to 8 cm, and detected PG events were analyzed within a 4.2-6.3 MeV energy window. System performance was evaluated using sensitivity (total detected counts) and spatial accuracy (Δ), defined as the absolute difference between true and reconstructed depth of maximum PG emission. The simulated energy spectrum exhibited characteristic photopeaks from carbon and oxygen, and a strong spatial correlation was confirmed between the distal edge of the Bragg peak and the PG emission maximum. Increasing collimator length decreased sensitivity but generally improved spatial accuracy. The optimal trade-off was achieved with a 7 cm collimator, yielding the best spatial accuracy of Δ = 2 mm. This study identifies 7 cm as the optimal collimator length for a parallel-hole PG camera, achieving spatial accuracy comparable to more complex slit- or pinhole-based systems. The findings provide a quantitative design framework and underscore the critical importance of balancing geometric resolution with sufficient count statistics for precise and timely clinical range verification.