Water suppressed proton nuclear magnetic resonance (1H NMR) spectroscopy of human plasma has been described as successful in detection of malignancy. We designed a prospective study to test the hypothesis that in vitro NMR spectroscopy has a high sensitivity for detecting early breast cancer. One hundred and thirty-five women were referred for breast biopsy due to abnormal mammograms. One hundred of these were recruited through a population-based mammography screening project. Sixty-nine of 135 women were found to have breast cancer and their average line width of the methyl and methylene resonance in the plasma were compared to those women who had a benign or normal histopathology in the biopsy and to the line width for 100 healthy subjects from the same population. The mean line width at a half-height of the methyl and methylene resonances of the serum lipoprotein lipids in the NMR spectrum did not differ appreciably between the groups. The line width correlated highly with the serum triglycerides, but correction for the level of triglycerides did not improve the diagnostic accuracy of the line width. Receiver-operating characteristic analysis revealed a sensitivity of 61% and a false positive rate of 43% at the most beneficial cut-off of line width (39.7 Hz). In vitro NMR spectroscopy in our hands was thus not a useful diagnostic tool in patients with early breast cancer.