Assessment of proton nuclear magnetic resonance spectroscopy for detection of malignancy

Water-suppressed proton nuclear magnetic resonance spectra were generated (by using 360 and 500 MHz systems) from human plasma and serum samples taken from 35 apparently healthy individuals, 52 patients with overt malignancies, and 37 patients with hypertriglyceridemia (triglycerides greater than 200 mg/dL or 2.26 mmol/L). The line widths from the lipoprotein-lipid methylene and methyl resonances at approximately 1.3 and 0.9 ppm were averaged by the method of Fossel et al. (N Engl J Med 1986;315:1369-76), but, contrary to their findings, we were unable to distinguish normal individuals from those with malignant tumors (e.g., mean +/- SD line width at 360 MHz: normal group = 32.9 +/- 3.6 Hz, malignant group = 28.3 +/- 4.9 Hz). The average line-width measurements (y), however, varied with the triglyceride content (x, mg/dL) of the plasma or serum as follows (logarithmic transformation of the data determined at 360 MHz and regression analysis): y = 110 (x-0.27). Data from both nonmalignant and malignant specimens fit this equation, the coefficient of correlation being -0.91. These findings suggest that considerable caution should be used in interpreting water-suppressed proton NMR spectra for cancer detection.