A Monte Carlo based intensity-modulated radiation therapy (IMRT) treatment planning system has been developed and used for breast treatment. An iterative method was used for optimization to generate IMRT plans and a step-and-shoot technique was used for beam delivery. The patient setup and incident beam directions were the same as those for conventional tangential photon treatment. The weights for the opposed beamlets in the two tangential beams were determined first by the doses at the depths of the maximum dose at both sides to minimize hot spots. The intensity of an individual beamlet pair was then optimized based on the dose at the midplane. Fine tuning was made to achieve optimal target dose uniformity and to reduce the dose to the heart when necessary. The final dose calculations were performed using the Monte Carlo method and the plans were verified by phantom measurements. The dose distributions and dose-volume-histograms of IMRT plans were compared with those of conventional plans that were generated using a commercial treatment planning system and recalculated using an in-house Monte Carlo system for the first 25 patients. The dose comparisons showed that the percentage volume receiving more than 95% of the prescription dose (V95) and the percentage volume receiving more than 100% of the prescription dose (V100) for the clinical target volume (CTV) of IMRT plans were about the same as those of conventional plans. The percentage volume receiving more than 105% of the prescription dose (V105) for the CTV was reduced from 23.1% to 7.9% on average. The percentage volume of the lung receiving more than 20 Gy dose (V20 Gy) during the entire treatment was reduced by about 10%. The percentage volume of the heart receiving more than 30 Gy dose (V30 Gy) is reduced from 3.3% to 0.3%. Further studies revealed that a less than 5 degrees change in couch angle and collimator angle at patient setup had no significant effect on the dose coverage of CTV but had significant effect on the dose to the lung and heart. The study on the effect of beam spoiler showed that it increased the dose at the buildup region by 0- 13% that varies with location. The machine output linearity and stability for small monitor unit delivery of Siemens accelerators used for this study was checked and found to be suitable for breast IMRT. The total effect of variations was calculated to be less than 1% for typical breast treatments. The beam delivery time was increased by about 2 min compared with conventional tangential treatments. The whole treatment including patient setup and beam delivery can be completed in a 15 min slot. The IMRT technique has been proven practical for breast treatment clinically. The results showed that tangential IMRT improved the dose homogeneity in the breast and reduced the dose to the lung and heart.