Detection of malignant tumors by nuclear magnetic resonance spectroscopy of plasma

Use of proton nuclear magnetic resonance spectroscopy of plasma in screening for malignant disease

In 1986, narrow plasma proton nuclear magnetic resonance (NMR) methyl and methylene line widths were reported to be associated with malignant disease, but more recent studies have not confirmed this relationship. The authors analyzed 106 plasma samples from healthy control subjects and patients with cirrhosis, hepatocellular carcinoma, metastatic liver tumors, other untreated cancers, and hyperlipidemia. NMR spectroscopy was performed using a proton NMR spectrometer operating at 399.65 MHz. A significant difference was found between the mean line widths of the plasma methyl resonances in control subjects and those in patients with cancer or hyperlipidemia. However, no significant difference was found between the mean plasma methylene line widths in control subjects and patients with cancer. Plasma samples from patients with liver disease or hyperlipidemia showed a characteristic methylene spectral pattern. The methylene pattern could be separated into three types: type A had a small peak on the right shoulder of the main peak; type B was a sharp single peak; and type C was a broad single peak. All control subjects had type A pattern; patients with liver disease had type C pattern; and patients who had hyperlipidemia had type B pattern, and hyperlipidemia may affect methyl and methylene line widths in NMR spectra. Because the methyl and methylene levels and their average line widths correlated inversely with triglyceride levels, considering the spectral patterns that indicate hyperlipidemia should decrease false-positive results and make the methyl line width useful for cancer screening.

Proton magnetic resonance spectroscopy of fractionated plasma lipoproteins and reconstituted plasma from healthy subjects and patients with cancer

Proton nuclear magnetic resonance spectra at 500 MHz of plasma and the very low density lipoprotein (VLDL), low density lipoprotein (LDL), and high density lipoprotein (HDL) fractions isolated by KBr gradient ultracentrifugation were analysed in 16 cancer patients, six pregnant and nine non-pregnant healthy subjects. In spectra with narrow plasma composite aliphatic peaks (methylene at 1.2-1.4 p.p.m. and methyl at 0.8-0.9 p.p.m., respectively), a relative increase in either VLDL, LDL, or both, or a decrease in HDL signals was observed. The mechanism for line-width narrowing seemed different in cancer patients (less signals from HDL relative to VLDL) compared with pregnant women (more signals from LDL). By reconstitution of plasma samples from both healthy subjects and patients with malignant disease, decreased concentration of VLDL or HDL resulted in broadening or narrowing of the composite peaks, respectively. The effects of VLDL and HDL on the plasma line width were moderated by the signals from LDL. Within lipoprotein fractions, the methylene and methyl resonances were shifted to a higher field with increased observation temperature, the change in shift being greatest for HDL. The line width of composite peaks in plasma varied with the observation temperature, depending on the relative concentrations of individual lipoproteins. The correlation coefficient (r) for the relation between total plasma triglyceride level and the average of the line-width of the composite methylene and methyl peaks was -0.78 (p less than 0.001). For spectra of individual lipoproteins, statistical significant relationships were found between line-widths and triglyceride content of the LDL fraction (methyl line-width, r = -0.63) (p less than 0.001) and between methylene line-width and cholesterol of HDL (r = 0.54) (p = 0.003). In summary, the shape and width of the composite aliphatic peaks of plasma were affected by the relative concentration, chemical shift and transition temperature of both VLDL, LDL, and HDL, and by the total triglyceride level. Comparing pregnancy and malignant disease, the lipoprotein resonances contributed differently in giving narrow composite signals.

Proton nuclear magnetic resonance lineshape studies on human blood plasma lipids from newborn infants, healthy adults, and adults with tumors

The usefulness of proton NMR spectroscopy of human blood plasma for cancer research has been extensively studied in recent years. Two main starting points have been offered by Fossel et al. (N. Engl. J. Med. 315, 1369 (1986)) and Mountford et al. (FEBS Lett. 203, 164 (1986)). In this work the experimental proton NMR spectra of blood plasma were analyzed with the aid of the multivariate lineshape fitting method. An appropriate model structure, in terms of the various lipoprotein (VLDL, LDL, and HDL) signals, for the methylene region was used. Neonates, healthy adults, and adults with nonmalignant and malignant tumors were studied. The linewidth of the methylene region was found to be linearly dependent on the relative concentrations of the lipoproteins. The correlation coefficient was -0.89 (P less than 0.001) for VLDL and 0.88 (P less than 0.001) for HDL. A correlation between VLDL concentration and age, 0.76 (P less than 0.001), was also established. VLDL was modeled using two components. The half-linewidth of the lower field component was slightly elevated for the adults with large metastases. This might be in association with the fucose-containing proteolipid complex detected earlier in cancer cells or in sera of cancer patients. Some signals of this complex may fall in the same region of the spectra. The spectra for the neonates were indicated to be totally different from the adults. This and other related questions were explained by means of the model parameters and the relative concentrations of the lipoproteins VLDL, LDL, and HDL. The presented technique can be used as a rapid research tool for figuring out the relative concentrations of the lipoproteins in blood plasma and explaining the reasons behind the changes in the spectra.

The proton NMR of blood plasma and the test for cancer

The water-suppressed proton NMR spectra of plasma from healthy controls and cancer patients have been recorded along with the high-density lipoprotein (HDL) and triglyceride (TG) levels in the plasma. The dependence of the average of methylene and methyl resonance linewidths on the HDL/TG ratio is demonstrated. It is shown that the deranged lipoprotein levels caused by the presence of cancer lead to the serological test for cancer proposed by Fossel. The work provides the formation for a complete understanding of the proposed serological test for cancer. However, the wide variation in lipoprotein composition in the plasma from both healthy adults and cancer patients means that the serological test proposed by Fossel cannot be used as a screen for cancer.

Proton nuclear magnetic resonance spectroscopy of plasma from healthy subjects and patients with cancer

To evaluate the ability of proton nuclear magnetic resonance (NMR) spectroscopy to indicate the presence or absence of malignant disease, we analyzed plasma samples from 104 patients with untreated cancer of various types and from 164 healthy controls. All specimens were coded with random numbers, and the investigators were blind to patient category. A statistically significant difference (P less than 0.001) was found between the mean (+/- SD) line widths in the plasma samples from the controls (39.1 +/- 6.7 Hz) and the line widths in plasma from the patients with cancer (35.2 +/- 6.4 Hz). However, the values showed considerable overlap between the two groups. The average line widths in the 54 male (36.0 +/- 7.9 Hz) and the 110 female (40.5 +/- 5.6 Hz) controls were significantly different (P less than 0.001). Differences in the average line width were also found between 34 male controls 40 years old or older (33.9 +/- 6.5 Hz) and 20 younger men (39.6 +/- 8.8 Hz) (P = 0.008) and between 61 female controls 40 or older (38.8 +/- 5.7 Hz) and 49 younger women (42.5 +/- 4.7 Hz) (P less than 0.001). The average line widths in 36 women with cancer (35.5 +/- 6.8 Hz) and their controls matched for age and sex (39.0 +/- 6.3 Hz) were significantly different (P = 0.03) but again showed much overlap. In 29 men with cancer, the line widths were not different from those of controls matched for age and sex. We conclude that proton NMR spectroscopy is not generally reliable for the detection of cancer. Furthermore, our data demonstrate the importance of studying control groups matched for age and sex.

Lack of efficacy of water-suppressed proton nuclear magnetic resonance spectroscopy of plasma for the detection of malignant tumors

Water-suppressed proton nuclear magnetic resonance (NMR) spectroscopy of plasma has been proposed by Fossel et al. (N Engl J Med 1986; 315:1369-76) as a technique for detecting malignant tumors. In their analysis, plasma samples from patients with cancer were clearly distinguished from those of normal subjects by measuring and averaging the methyl and methylene line widths of plasma lipoproteins in NMR spectrums. To evaluate this diagnostic procedure further, we collected and analyzed by NMR spectroscopy 145 samples of plasma from patients who served as controls, most of whom were undergoing orthopedic or cardiac surgery (n = 66); patients with a variety of untreated malignant tumors (n = 25) or treated malignant tumors (n = 18); and patients with hyperplastic or "premalignant" diseases, such as benign prostatic hyperplasia and ulcerative colitis (n = 36). All the samples were coded, and NMR spectroscopy was performed without knowledge of the patients' clinical status. There were no significant differences in the NMR line widths among the four study groups (P greater than 0.05 for all pairwise comparisons). The specificity and sensitivity of this method for distinguishing the control patients (mean line width [+/- SD], 44.0 +/- 7.4 Hz) from those with untreated cancer (43.8 +/- 6.9 Hz) were poor, with a false positive rate of 52 percent (34 of 66) and a false negative rate of 56 percent (14 of 25). Inverse correlations of line widths with age (P less than 0.01) and with the plasma triglyceride level (P less than 0.001) were detected. We conclude that NMR spectroscopy of plasma is not an accurate test for the detection of malignant tumors.

Investigations concerning the potential for using 1H NMR relaxometry or high-resolution spectroscopy of plasma as a screening test for malignant lung disease

In 158 plasma samples, obtained from patients with lung carcinoma, lung metastases, and infectious or inflammatory lung diseases and from healthy controls, the NMR relaxation times T1 and T2 of water protons were measured at a resonance frequency of 20 MHz by pulsed NMR techniques and adjusted to a standardized total plasma protein concentration. For one-third of these samples water-suppressed 500-MHz 1H NMR spectroscopy at 37 degrees C was used (a) to determine the widths of the composite lipid methyl and methylene signals, and (b) to quantitate individual lipid methylene signal components that could be detected in resolution-enhanced spectra. In addition, hematological parameters and the plasma levels of several acute phase proteins and apolipoprotein-A were monitored. No diagnostically significant differences between lung carcinoma patients and patients with nonmalignant lung disease could be found for any of the plasma NMR parameters, nor could T1 or lipid linewidth data distinguish between any patient group and healthy controls. However, the mean T2 was significantly shortened by about 15% for any kind of lung disease compared to healthy controls. Similar but less significant results were found for apolipoprotein-A levels. A linear discriminant function, calculated from the apolipoprotein-A and T2 data, did not improve the differentiation between malignant and nonmalignant lung disease but did improve the discrimination between tumor patients and healthy controls up to a sensitivity and specificity of 80 and 96.5%, respectively. T2 correlates inversely with plasma fibrinogen levels and the blood sedimentation rate and, therefore, appears to monitor a general inflammatory status of a tumor patient rather than the presence or absence of cancer. For all groups except healthy pregnant women, the lipid methylene composite signal linewidth correlates inversely with the fraction of mobile triglyceride present (mainly as VLDL), as estimated from resolution-enhanced spectra.

NMR spectroscopy evaluation of plasma "oncolipids" in head and neck cancer

The use of water-suppressed proton nuclear magnetic resonance spectroscopy of plasma as a serologic test for the detection of malignancy was first described in 1986. That report prompted the present study, which was undertaken to evaluate the efficacy of this test in differentiating patients who have head and neck malignancy from normal controls. Forty-six patients who had a biopsy-proven malignancy of the head and neck and 32 healthy individuals provided plasma for which the nuclear magnetic resonance spectrum was plotted, blind to patient diagnosis or group. The average line-width of methyl and methylene resonance was calculated. Significant differences (p less than 0.05) were found between the group with disease and the group with no disease for the methyl line-widths, using analysis of variance. In spite of this statistical difference, plotting of the values for the methyl, methylene, and average line-widths clearly demonstrated that these three oncolipid measures have no clinical use because of the tremendous overlap between the disease and nondisease groups. The findings of this study do not support the use of water-suppressed proton nuclear magnetic resonance spectroscopy as a clinically useful test for the diagnosis of head and neck malignancy.

Alteration of aliphatic lipid proton NMR linewidths by malignant tumors in guinea pigs

Water-suppressed proton nuclear magnetic resonance spectroscopy was used to observe plasma lipoprotein lipid methyl and methylene resonances from guinea pigs which had been injected with viable or heat-killed line 1 or line 10 tumor cells or sterile oil. It was shown that the widths of these resonances became significantly sharper as the number of tumor cells grew. Plasma from tumor-free control animals showed no change in the NMR linewidths. It is concluded that the changes observed reflect a specific host response to viable tumor cells, and in these models there is a reciprocal relationship between the number of viable tumor cells and the linewidths of plasma lipoprotein methyl and methylene resonances.

Nuclear magnetic resonance spectroscopy of plasma for the detection of head and neck cancer

The plasma of 15 normal volunteers, four patients with benign head and neck disease, and 12 patients with biopsy proven head and neck malignancies were evaluated using water-suppressed nuclear magnetic resonance (NMR) spectroscopy. While the mean full width at half height (HHLW) of the NMR spectra showed a significant difference between groups (P less than .05), the predictive value of a positive test was only 57% the sensitivity was only 33%, and individual values in all groups demonstrated considerable overlap. The mean HHLW of the control group was 38.5 Hz +/- 3.5 compared with 35.2 Hz +/- 4.7 for the cancer group and 35.0 Hz +/- 10 for the benign disease group. This method was not able to distinguish patients with malignancy from those with benign disease or controls. Its use as a specific screening method for head and neck malignancy cannot be recommended.

Detection of tumours with nuclear magnetic resonance spectroscopy of plasma

A systematic investigation of cancer detection by water-suppressed nuclear magnetic resonance (NMR) of plasma is reported. With additional suppression of lactate, a statistically significant difference between the linewidths of the methylene group signal of patients with untreated cancer (average linewidth 26.9 +/- 3.9 Hz) and normal controls (average linewidth 31.1 +/- 4.9 Hz) has been found. However, overlap was found between these two groups. It is shown that recognition of malignancy could be improved by consideration of the different relations of the linewidths on the content of serum triglycerides and the observation of a shoulder at the high field side of the methylene signal. Preliminary investigations on lipid fractions separated by ultracentrifugation (UC) indicate a connection of the appearance of the high field shoulder and the HDL lipoprotein.

1H and 13C nuclear magnetic resonance studies of plasma from patients with primary intracranial neoplasms

Plasma from patients with malignant and benign primary intracranial neoplasms and from healthy control subjects has been examined using proton (1H) and carbon-13 nuclear magnetic resonance (NMR) spectroscopy. Some features in the spectra of plasma from patients with malignant tumors differ significantly (p less than 0.01) from the corresponding features in the spectra of plasma from control subjects and from patients with benign tumors. The NMR spectral parameters vary consistently with the Kernohan grade of astrocytoma, which may suggest that they give a measure of tumor growth kinetics. The observed spectral differences are shown to be due to elevated levels of monounsaturated and polyunsaturated fatty acid residues in the plasma of cancer patients. It is proposed that these lipid residues arise from cell membranes shed from cells in growing tumors. The ability to follow tumor growth kinetics directly may be of considerable importance in elucidating the effects of primary intracranial neoplasm therapy.

Detection of malignant tumors: water-suppressed proton nuclear magnetic resonance spectroscopy of plasma

Average linewidths of methyl and methylene proton resonances of plasma lipoproteins from healthy controls and untreated cancer patients were measured in water-suppressed proton spectra obtained at 400 MHz. In contrast to a previous report (E. T. Fossel, J. M. Carr, and J. McDonagh, N. Engl. J. Med. 315, 1369 (1986] no statistically significant difference (P greater than 0.5) was found between the average linewidths in plasma samples from 20 normal controls (36.0 +/- 6.4 Hz) and 17 untreated cancer patients (34.6 +/- 6.3 Hz). When the effect of lactate was removed by graphical extrapolation, a small but marginally significant (P = 0.08) difference was observed between the average linewidths of the normal controls (35.1 +/- 5.8 Hz) and untreated cancer patients (31.8 +/- 5.2 Hz). However, there was still a large overlap between the two groups. In addition, substantial day to day variations in the linewidths of samples from the same individuals were observed. We conclude that these measurements cannot be used to detect cancer.