Various approaches to reduce dose to the female breast in thoracic CT have been investigated. We evaluated the potential for reduction of dose to the breast by optimal choice of the X-ray spectra. The effect of X-ray energy variation on dose to the female breast in thoracic CT was examined by simulations and measurements of image contrast, image noise and radiation dose. A standard thorax phantom was used with various extension rings and breasts added and the following contrast inserts: iodine, calcium hydroxyapatite and a pure soft-tissue density difference. Three-dimensional dose distributions were determined by a validated Monte Carlo tool. The contrast-to-noise ratio per unit dose (CNRD) was determined for tube voltages of 40-200 kV by simulations and for 60-140 kV by measurements on a clinical CT scanner. CNRD curves did not show significant variations in soft-tissue density contrast, but considerable optimisation potential for iodine and skeletal imaging at reduced energies. Exact values depend on the patient's cross-section and X-ray spectrum. For example, reducing the tube voltage from 120 kV to 80 kV on the scanner reduced dose to the female breast typically by 50% without deterioration of the CNR. This method exceeds the dose reduction potential of other measures. We conclude that tube voltages in thoracic CT can be lowered for contrast medium and skeletal imaging without affecting the CNR but with a significant decrease in dose to the female breast.