A vaginal suppository containing ulinastatin (UTI) was developed as a hospital pharmacy product from UTI injection solution and Witepsol^® S-55. After mixing at 50°C for 0–8 h, UTI suppositories were prepared, which had good UTI content uniformity. Because 2% surfactant was contained in S-55, the UTI injection solution formed a water-in-oil type emulsion as a suppository base. The measured residual moisture content (loss on drying (LOD)) in the prepared vaginal suppositories decreased as the mixing time increased, but their hardness (hardness test (HT)) increased. Near (N) IR spectra of UTI suppositories were measured after mixing for 0–8 h. The best calibration models to predict the HT and LOD of the suppositories were determined based on the NIR spectra by the leave-one-out method in a partial least-squares regression analysis (PLS). The validation result indicated that PLS models for HT and LOD were obtained based on the spectra treated by a combination of smoothing and normalized, respectively, and the model consisted of three latent variables. The plots between the predicted and measured pharmaceutical properties (HT and LOD) based on the calibration data were superimposed with those of the external validation data. The developed NIR spectroscopy method was applied to the preparation process monitoring for UTI vaginal suppositories. In the prepared vaginal suppositories, the predicted LOD decreased as the mixing time increased, and the measured LOD values superimposed well with the predicted values. In contrast, the predicted HT increased as the mixing time increased, and the measured values superimposed with the predicted values.