Vegetated buffers are a well-studied and widely used agricultural management practice for reducing nonpoint-source pollution. A wealth of literature provides experimental data on their mitigation efficacy. This paper aggregated many of these results and performed a meta-analysis to quantify the relationships between pollutant removal efficacy and buffer width, buffer slope, soil type, and vegetation type. Theoretical models for removal efficacy (Y) vs. buffer width (w) were derived and tested against data from the surveyed literature using statistical analyses. A model of the form Y = K x (1-e(-bxw)), (0 < K < or = 100) successfully captured the relationship between buffer width and pollutant removal, where K reflects the maximum removal efficacy of the buffer and b reflects its probability to remove any single particle of pollutant in a unit distance. Buffer width alone explains 37, 60, 44, and 35% of the total variance in removal efficacy for sediment, pesticides, N, and P, respectively. Buffer slope was linearly associated with sediment removal efficacy either positively (when slope < or = 10%) or negatively (when slope > 10%). Buffers composed of trees have higher N and P removal efficacy than buffers composed of grasses or mixtures of grasses and trees. Soil drainage type did not show a significant effect on pollutant removal efficacy. Based on our analysis, a 30-m buffer under favorable slope conditions (approximately 10%) removes more than 85% of all the studied pollutants. These models predicting optimal buffer width/slope can be instrumental in the design, implementation, and modeling of vegetated buffers for treating agricultural runoff.