A method for analysis of silicon in tissue was developed to determine silicon content in breast parenchymal and periprosthetic capsular tissues of patients with silicone or saline implants and to compare levels in tissues from normal (nonaugmented) breasts. It is of interest to determine whether increased silicon content in tissues can be associated with morbidity in patients who have received silicone implants. This manuscript addresses the issues involved in analysis of breast tissue samples for silicon and compares silicon levels with tissue histologic findings and patient morbidity. One hundred sixty tissue samples were obtained for silicon analysis from 72 patients during augmentation, capsulectomy with or without replacement mammoplasty, mastectomy, or biopsy procedures and were frozen in acid-washed polystyrene tubes at 220 degrees C until analysis. Samples were thawed, sectioned to approximately 0.1 g (dry weight), and digested in nitric acid before analysis by inductively coupled plasma emission spectroscopy, monitoring emission intensity at 251.6 nm. Tissue silicon levels (breast parenchymal and periprosthetic capsular tissue) in patients with silicone gel implants were much higher (mean, 9,287 micrograms/g, n = 106) than in patients with saline implants (mean, 196 micrograms/g, n = 37) or nonaugmented breasts (mean, 64 micrograms/g, n = 17). Histologic examination was performed on 54 tissue samples stained with hematoxylin-eosin. Tissue samples were rated as to degree of inflammation and calcification, and amount of giant cells, foamy histiocytes, and vacuoles containing a colorless refractory material. Vacuolization and foamy histiocyte ratings correlated significantly with tissue silicon concentration. No correlations were found between tissue silicon concentration and inflammation, calcification, or giant cell rating. Implant age (number of years an implant was in place before sampling) correlated with capsular tissue silicon concentration in patients with intact implants but not in those with ruptured implants. No difference in tissue silicon concentration was found between patients with or without signs or symptoms of morbidity. Using 0.1 g of tissue, the method was linear to 1,000 micrograms/g, and sensitivity was 3.7 micrograms/g. Precision between runs (mean, 5.1 micrograms/g; coefficient of variance, 13.7%; n = 13) was calculated from multiple analyses of a bovine liver standard (National Bureau of Standards, reference material 1577a). Significant biologic variability (21.4% to 52.5%) was seen in tissues with high silicon levels. Paraffin-embedded, formalin-fixed tissues are not amenable to silicon analysis by this method, because of leaching of silicone from the tissues during preparation. Thus only fresh frozen tissue samples were used.