Water-suppressed proton nuclear magnetic resonance (NMR) of plasma was proposed as a technique for detecting malignant tumors. In that analysis, bloods drawn from cancer patients at the Beth Israel Hospital (BIH; Boston, MA), were easily distinguished from normal subjects by measuring and averaging the proton NMR methyl and methylene line widths of plasma lipoproteins. We collected blood at the Massachusetts General Hospital (MGH), including from normal controls, patients with untreated and treated malignant tumors, and patients with nontumor diseases. The plasma NMR analyses were carried out blind. The code was not broken until all patient charts and pathology records were reviewed, plasma analyses were completed, and patients had been divided into appropriate clinical groups. Analysis of these data showed no differences between the means of the study groups (false-positive and false-negative frequencies 46% and 57%, respectively). An inverse correlation of methyl/methylene line widths with age (P less than .01), and a correlation with nitrate-requiring cardiovascular disease (P less than .05) was, however, evident. This test cannot be validly used to detect malignancy.