Purpose The aim of this study was to determine the role that frame and lens design play in lens retention during high-impact testing of safety eyewear that advertises conformance to the performance-based ANSI Z87.1-2003 standard. Methods A total of 75 Z87 safety eyeglass frames (3 each of 25 frame models) were used in this study, procured from 5 of the leading U.S. safety frame manufacturers. Frames were fitted by an independent laboratory with 2.0-mm plano polycarbonate lenses in compliance with ANSI Z87.1-2003. Finished spectacles were sent to a subsequent laboratory testing facility where each frame was subjected to both high-mass and oblique-incidence high-velocity impacts to determine frame characteristics that were most highly associated with testing failure. Among the frame and lens parameters that were considered in this analysis were the A and B dimensions, effective diameter, distance between lenses, bridge type, frame material, bevel type, and frame cost. Certain variables were controlled for by maintaining consistency among all spectacle pairs, e.g., lens prescription, center thickness, and edge thickness. Multiple logistic regression was used to control potential confounding variables and to develop the best combination of them for predictive value. Results Of 25 separate frame models assessed, 10 passed both high-mass and high-velocity impact testing, i.e., none of the 3 frame/lens samples failed. Of the models that failed, 13 failures were caused by high-velocity testing, 1 by high-mass testing, and 1 failed both high-mass and high-velocity testing. None of the 15 spectacles with the SprinGuard™ (Hilco, Plainville, Massachusetts) bevel design failed, although these were proprietary to 1 manufacturer and included only 5 frame models. Two spectacle designs (6 individual frames) incorporated an inverted bevel design of which 3 of the frames failed impact testing. Controlling for drop ball velocity among the 54 remaining standard “V” bevel spectacle pairs, the odds of failure were about 8 to 9 times higher for metal frames than Zylonite frames (odds ratio [OR], 8.5; 95% confidence interval [CI], 1.4 to 52.3; P = 0.02), and the odds of failure were about 4 times higher for lens effective diameters of less than 50 mm than for lens effective diameters more than 50 mm (OR, 4.1; 95% CI, 1.2 to 14.8; P = 0.03). Conclusion Among our sample of safety spectacles, failure from high-mass impact resistance testing rarely occurred. No spectacles with the SprinGuard bevel design failed in our analysis, but this was limited to a small sample size from a single manufacturer. Among our sample of 54 spectacle pairs with a v-bevel design, metal frame material and an effective diameter of less than 50 mm were the strongest predictors of failure relative to high-velocity impact testing with a 45° temporal angle projectile.