Fossel et al. have recently proposed the proton NMR examination of plasmatic lipoproteins--and more precisely the determination of an index obtained from the averaged linewidth of the CH2 and CH3 resonances--as a possible tool for detection of cancer. Many evaluations conducted on an international basis have demonstrated that initial expectations were not met and that the test lacked sensitivity, specificity, and predictive value to be accepted as a screening and diagnostic tool. In our evaluation we have collected plasma from healthy subjects, from patients with various kinds of cancer at different stages of evolution and therapy, and from patients suffering from a variety of pathologies, including benign tumors. In accordance with Chmurny et al., we observed that the linewidth index (LWI) is precise and reproducible when care is taken in the handling and storage of samples and in the fasting of subjects. After finding no predictive value to the test, we have reanalyzed the spectra and studied the variations of the ratio defined by the methylene signal area over the methyl signal area. This ratio is significantly increased in cancer. Furthermore, it offers a better separation of statistical populations permitting a more precise discrimination between cancer, other pathologies and controls. We have also found that malignant tumors arising from mesenchyma (sarcoma, leukemia, lymphoma) induce less important variations in the CH2/CH3 ratio than adenocarcinoma or glioma, when such differences cannot be documented using the LWI. These observations are particularly interesting since they might bring new information on the metabolic modifications of the LWI and the CH2/CH3 ratio might reflect the embryologic origin of the tumors and raise the issue of the heterogeneity of cancer disease.