LETTERS
Electroconvulsive Therapy and Brain Lipids
Reffences
John A. Detre, M D In their recent report Woods and Chiu 111 set out to use in vivo ‘H magnetic resonance spectroscopy (MRS) to study alterations in brain lactate produced by electroconvulsive therapy (ECT). Instead, the authors concluded that ECT treatments produce alterations in “brain lipids” based on difference spectra obtained before and after ECT therapy. It is surprising that no lactate resonance was demonstrated anywhere in the paper. Several authors have successfully obtained lactate measurements from human brain using ‘H MRS 12-41; however, doing so requires sufficient spatial localization to exclude contamination from subcutaneous and cranial fat. Figure 2 in the article by Woods and Chiu shows two spectra obtained before and after ECT (B and C). Both contain broad lipid resonances which appear 90 degrees out of phase with the NAA resonance. Since it is widely accepted that, at least in normal brain, little or no fat resonance is detected using ‘H MRS, it is likely that the observed lipid signal is due to contamination from outside the VOI. This lipid signal obscures the underlying lactate resonance. Some mention is made in the paper of lactate editing experiments, but these results are not shown. The authors attempt to demonstrate an increase in lactate by subtracting spectra obtained before and after ECT. While difference spectroscopy is a commonly used technique, it is valid only when both spectra are obtained under identical conditions, generally in immediate succession. Since it would be impossible to perform ECT inside a high-field magnet, these conditions clearly cannot be met. Small variations in coil tuning, loading, or magnetic field homogeneity could easily account for the differences in fat contamination observed. The authors exclude fat contamination by stating that the lipid signal increased in 5 successive patients. Although this may have been the case, in Figures 2, 3, and 4 “difference” spectra are shown with random phasing of the NAA resonance, making it difficult to determine whether the difference was, in fact, positive or negative. Since the NAA resonance was used for phasing the spectra, it is hard to understand why the phase should not be the same in all spectra. Furthermore, the areas under the lipid peaks in the difference spectra appear t o be an order of magnitude greater than the NAA, suggesting a “brain lipid” concentration of 100 mM, several orders of magnitude greater than free fatty acid concentrations measured in mammalian brain after bicuculline-induced seizures 153. While in vivo ‘H MRS offers a potentially powerful window into human brain metabolism, it is technically demanding and susceptible to artifact. The data presented by Woods and Chiu fail to address their original question and do not clearly support their conclusions. Department of Nearology Hospital of the Univenity of Penruylvania
Philadehhiu, P A
1. Woods BT, Chiu T-M. In vivo ‘H spectroscopy of the human brain following electroconvulsive therapy. Ann Neurol 1990; 28:745-749 2. Bruhn H, FrahmJ, Gyngell ML, et al. Noninvasive differentiation of tumors with use of localized H-1 MR spectroscopy in vivo: initial experience in patients with cerebral tumors. Radiology 19900;172:541-548 3. Hanstock CC, Rothman DL, Prichard JW, et al. Spatially localized ‘H NMR spectra of metabolites in the human brain. Proc Natl Acad Sci USA 1988;85:1821-1825 4. Detre JA, Wang 2, Bogdan AR, et al. Regional variation in brain lactate in Leigh syndrome by localized ‘H magnetic resonance spectroscopy. Ann Neurol 1991;29:218-221 5. Siesjo BK, Ingvar M, Westerberg E. The influence of bicuculline-induced seizures on free fatty acid concentrations in cerebral cortex, hippocampus, and cerebellum. J Neurochem 1982;39:796-802
Reply Bryan T. Woods, MD, and Tak-Ming Chin, PhD In response to Dr Detre’s objection rhar no lactate resonance was demonstrated, we did see a small lactate signal in 2 patients after electroconvulsive therapy (ECT), and have so noted in our paper {If. Whereas it is generally accepted that one sees little or no lipid signal in normal brain at long T E s , a readily identifiable lipid signal has been detected in normal brain at the short TE we used in our studies E2). In any case, we do not consider that human brains studied 1 hour after ECT are “normal,” and the whole point of our paper was that the major effect of ECT was a marked quantitative increase in the brain lipid signal. As to the likelihood of inadvertent contamination, it would have to take place on only those scans done after ECT, and never on the ones done before, and we have now made this same lipid signal observation in 8 separate patients. Of a total of 43 separate scans, 17 of which were done within 90 minutes of ECT, and 26 done either before or more than 36 hours after ECT, we have no failures of the lipid signal to increase shortly after ECT in comparison to control studies in the same patients. Dr Detre seems to misunderstand the subtraction technique. It is well known that in difference spectroscopy {3], a residual peak resembling a derivative signal may result from subtraction of two peaks with identical amplitudes but slightly different phase and frequency. Subtraction errors of this kind may be due to spectrometer stability and are relatively common. In any case, incomplete subtraction of the N-AA peak before and after ECT would not, in itself, generate a lipid signal at a different position in the ‘H spectrum. Furthermore, one should not compare the area under the lipid peak to the area under the N-AA peak in the difference spectra, as Dr Detre suggests, because the N-AA peak is in large part subtracted out. We indicated in our paper [1f that in 1 of the studies the maximum lipid to N-AA area ratio was 3.75. In interpreting this it should be recalled that the area under a magnetic resonance spectroscopy (MRS) peak is determined by the number of molecules times the number of protons per molecule. Thus free fatty acids (FFA) are likely to have 30 to 40 and diacylglycerol (DAG) 60 to 80 resonant protons per
Copyright 0 1991 by the American Neurological Association 429
Electroconvulsive Therapy and Brain Lipids
Reffences
John A. Detre, M D In their recent report Woods and Chiu 111 set out to use in vivo ‘H magnetic resonance spectroscopy (MRS) to study alterations in brain lactate produced by electroconvulsive therapy (ECT). Instead, the authors concluded that ECT treatments produce alterations in “brain lipids” based on difference spectra obtained before and after ECT therapy. It is surprising that no lactate resonance was demonstrated anywhere in the paper. Several authors have successfully obtained lactate measurements from human brain using ‘H MRS 12-41; however, doing so requires sufficient spatial localization to exclude contamination from subcutaneous and cranial fat. Figure 2 in the article by Woods and Chiu shows two spectra obtained before and after ECT (B and C). Both contain broad lipid resonances which appear 90 degrees out of phase with the NAA resonance. Since it is widely accepted that, at least in normal brain, little or no fat resonance is detected using ‘H MRS, it is likely that the observed lipid signal is due to contamination from outside the VOI. This lipid signal obscures the underlying lactate resonance. Some mention is made in the paper of lactate editing experiments, but these results are not shown. The authors attempt to demonstrate an increase in lactate by subtracting spectra obtained before and after ECT. While difference spectroscopy is a commonly used technique, it is valid only when both spectra are obtained under identical conditions, generally in immediate succession. Since it would be impossible to perform ECT inside a high-field magnet, these conditions clearly cannot be met. Small variations in coil tuning, loading, or magnetic field homogeneity could easily account for the differences in fat contamination observed. The authors exclude fat contamination by stating that the lipid signal increased in 5 successive patients. Although this may have been the case, in Figures 2, 3, and 4 “difference” spectra are shown with random phasing of the NAA resonance, making it difficult to determine whether the difference was, in fact, positive or negative. Since the NAA resonance was used for phasing the spectra, it is hard to understand why the phase should not be the same in all spectra. Furthermore, the areas under the lipid peaks in the difference spectra appear t o be an order of magnitude greater than the NAA, suggesting a “brain lipid” concentration of 100 mM, several orders of magnitude greater than free fatty acid concentrations measured in mammalian brain after bicuculline-induced seizures 153. While in vivo ‘H MRS offers a potentially powerful window into human brain metabolism, it is technically demanding and susceptible to artifact. The data presented by Woods and Chiu fail to address their original question and do not clearly support their conclusions. Department of Nearology Hospital of the Univenity of Penruylvania
Philadehhiu, P A
1. Woods BT, Chiu T-M. In vivo ‘H spectroscopy of the human brain following electroconvulsive therapy. Ann Neurol 1990; 28:745-749 2. Bruhn H, FrahmJ, Gyngell ML, et al. Noninvasive differentiation of tumors with use of localized H-1 MR spectroscopy in vivo: initial experience in patients with cerebral tumors. Radiology 19900;172:541-548 3. Hanstock CC, Rothman DL, Prichard JW, et al. Spatially localized ‘H NMR spectra of metabolites in the human brain. Proc Natl Acad Sci USA 1988;85:1821-1825 4. Detre JA, Wang 2, Bogdan AR, et al. Regional variation in brain lactate in Leigh syndrome by localized ‘H magnetic resonance spectroscopy. Ann Neurol 1991;29:218-221 5. Siesjo BK, Ingvar M, Westerberg E. The influence of bicuculline-induced seizures on free fatty acid concentrations in cerebral cortex, hippocampus, and cerebellum. J Neurochem 1982;39:796-802
Reply Bryan T. Woods, MD, and Tak-Ming Chin, PhD In response to Dr Detre’s objection rhar no lactate resonance was demonstrated, we did see a small lactate signal in 2 patients after electroconvulsive therapy (ECT), and have so noted in our paper {If. Whereas it is generally accepted that one sees little or no lipid signal in normal brain at long T E s , a readily identifiable lipid signal has been detected in normal brain at the short TE we used in our studies E2). In any case, we do not consider that human brains studied 1 hour after ECT are “normal,” and the whole point of our paper was that the major effect of ECT was a marked quantitative increase in the brain lipid signal. As to the likelihood of inadvertent contamination, it would have to take place on only those scans done after ECT, and never on the ones done before, and we have now made this same lipid signal observation in 8 separate patients. Of a total of 43 separate scans, 17 of which were done within 90 minutes of ECT, and 26 done either before or more than 36 hours after ECT, we have no failures of the lipid signal to increase shortly after ECT in comparison to control studies in the same patients. Dr Detre seems to misunderstand the subtraction technique. It is well known that in difference spectroscopy {3], a residual peak resembling a derivative signal may result from subtraction of two peaks with identical amplitudes but slightly different phase and frequency. Subtraction errors of this kind may be due to spectrometer stability and are relatively common. In any case, incomplete subtraction of the N-AA peak before and after ECT would not, in itself, generate a lipid signal at a different position in the ‘H spectrum. Furthermore, one should not compare the area under the lipid peak to the area under the N-AA peak in the difference spectra, as Dr Detre suggests, because the N-AA peak is in large part subtracted out. We indicated in our paper [1f that in 1 of the studies the maximum lipid to N-AA area ratio was 3.75. In interpreting this it should be recalled that the area under a magnetic resonance spectroscopy (MRS) peak is determined by the number of molecules times the number of protons per molecule. Thus free fatty acids (FFA) are likely to have 30 to 40 and diacylglycerol (DAG) 60 to 80 resonant protons per
Copyright 0 1991 by the American Neurological Association 429
