THE EEG EFFECTS OF ECT: IMPLICATIONS FOR rTMS ANDREW D. KRYSTAL*, MIKE WEST$, RAQUEL PRADO$, HENRY GREENSIDE@, SCOTT ZOLDI@, AND RICHARD D. WEINER* * Quantitative EEG Laboratory, ECT Program, $ Institute of Statistics and Decision Sciences, @ Center for Nonlinear and Complex Systems, Duke University, Durham, NC. To appear in: "Progress in Neuropsychopharmacology and Biological Psychiatry" Address all correspondence to: Andrew D. Krystal, PhD, MD Box 3309 Duke, University Medical Center, Durham, NC 27710. Phone: 919-681-8742. Fax: 919-681-8744 E-mail: krystal@phy.duke.edu ABSTRACT Electroconvulsive therapy (ECT) involves the use of electrical stimulation to elicit a series of generalized tonic-clonic seizures for therapeutic purposes and is the most effective treatment known for major depression. These treatments have significant neurophysiologic effects, many of which are manifest in the electroencephalogram (EEG). The relationship between EEG data and the response to electroconvulsive therapy (ECT) has been studied since the 1940's but for many years no consistent correlates were found. Recent studies indicate that a number of specific EEG features recorded during the induced seizures (ictal EEG) as well as before and after a course of treatment (inter-ictal EEG) are related to both the therapeutic efficacy and cognitive side-effects. Similar to ECT, repetitive transcranial magnetic stimulation (rTMS), which involves focal electromagnetic stimulation of cortical neurons, has also been studied as an antidepressant therapy and also appears to have neurophysiologic effects, although these have not been as fully investigated as is the case with ECT. Given the similarity of these treatments it is natural to consider whether advances in understanding the electrophysiologic correlates of the ECT response might have implications for rTMS. The present paper reviews the literature on the EEG effects of ECT and discusses the implications in terms of the likely efficacy and side-effects associated with rTMS in specific anatomic locations, the potential for producing an antidepressant response with rTMS without eliciting seizure activity, eliciting focal seizures with rTMS, and the possibility of using rTMS to focally modulate seizure induction and spread with ECT to optimize treatment Key Words: ECT, EEG, rTMS, Major Depression INTRODUCTION Electroconvulsive therapy (ECT), developed in the late 1930's, involves the use of electrical stimulation to elicit a series of generalized tonic-clonic seizures for therapeutic purposes, most commonly for the treatment of major depression (APA, 1990, Krystal and Weiner, 1994). Despite the development of antidepressant medications, beginning in the 1950's, ECT remains the most effective treatment known for major depression (APA, 1990, Krystal and Weiner, 1994). That ECT affects cerebral neurophysiology has been apparent since very early in its history. Reports of EEG recordings relating to ECT date back to as early as 1939, and studies examining the relationship between ECT response and EEG data obtained both during the induced seizures (ictal EEG) and during waking prior to and following the ECT course (inter-ictal EEG) have been carried out since the 1940's (Fleming et al., 1939, Proctor and Goodwin, 1943, Honcke and Zahle, 1946). While the early literature was unclear on the relationship of EEG indices to ECT outcome (Weiner, 1983), more recent studies suggest that consistent relationships exist (Nobler et al., 1993, Krystal et al., 1995, 1996, 1997, 1998, Sackeim et al., 1996, Roemer et al., 1996, Suppes et al., 1996, Krystal, 1998, Krystal and Weiner, In Press). Similarly to ECT, repetitive transcranial magnetic stimulation (rTMS) has also been studied as an antidepressant therapy and involves electromagnetic stimulation of the brain which appears to have neurophysiologic effects as well (Izumi et al., 1997, Wassermann, 1998). Given the commonalities of these treatments it is natural to consider whether recent advances in understanding the electrophysiologic correlates of the ECT response might have implications for rTMS. In the following reveiw, we consider recent studies of the relationship between EEG data and both the efficacy and cognitive side-effects of ECT. We then discuss the implications of these findings for the efficacy and side-effects of rTMS. EFFECTS OF ECT ON THE ICTAL EEG A. Basic Properties of the Ictal EEG The hallmark of effective electroconvulsive therapy (ECT) is the elicitation of a generalized tonic- clonic seizure, the occurrence of which is manifest in electroencephalographic (EEG) data recorded at the time of the treatments (Weiner et al., 1991, Krystal and Weiner, In Press). The capability to record ictal EEG data is now integrated into most present U.S. ECT devices and its use is now recommended for clinical practice in the USA (APA, 1990, Weiner et al., 1991, Krystal and Weiner, In Press). When recorded using prefrontal to ipsilateral ear or mastoid placement of electrodes (a common practice (Weiner et al., 1991, Krystal and Weiner, 1993, 1994, In Press, Weiner and Krystal, 1993)), a suprathreshold ECT stimulus leads to a seizure that is characterized by a typical sequence of EEG changes. While there is great variation in the manifestation of these effects, it has been possible to identify an idealized sequence of stages through which the ictal EEG typically progresses (Weiner et al., 1991, Weiner and Krystal, 1993, Krystal and Weiner, In Press). Immediately after the stimulus there is often a very brief period (up to several seconds) of low amplitude high-frequency activity, reflecting an initial desynchronizing effect of the electrical stimulus. This phase may be followed by an approximately 10 Hz low amplitude and regular rhythm referred to as the "epileptic recruiting rhythm" that is believed to be associated with an initial ictal syncrhonization. More consistently present folowing the stimulus, More consistently present folowing the stimulus, and coinciding with tonic and sometimes early clonic motor activity, is an ictal EEG phase characterized by 6-12 Hz poly-spike activity, often intermixed with occasional irregularly patterned slower frequency waveforms. With the transition from tonic to clonic motor activity, there is typically an increase in the stereotypy of waveform shape (generally termed morphologic regularity) and amplitude and a decrease in EEG frequency content. Frequently these changes evolve into a high amplitude pattern of slow-waves (initially in the 4-6 Hz range) accompanied by synchronous bursts of spike activity. This has been termed the "polyspike-and-slow wave" phase of the seizure. As the seizure progresses, the dominant frequency (the highest amplitude peak of the power spectrum) of the slow-waves decreases and may reach as low as 1-2 Hz. Towards the end of the seizure there may be a transitional period during which the poly-spike and slow wave pattern becomes less regular and diminishes in amplitude. The EEG manifestations of this termination phase of the seizure are extremely variable, ranging from a very gradual decline in frequency and amplitude to an abrupt termination of seizure activity. The termination of the seizure is followed by a relative suppression of EEG activity heralding the onset of the post- ictal period. The motor manifestations of the induced seizure terminate at a variable time with respect to the termination of ictal EEG changes, although never of longer duration (Weiner et al., 1991, Weiner and Krystal, 1993). The temporo-spatial characteristics of ECT seizures have been of great interest. It should be evident from the above description that one of the hallmarks of electrically induced seizures is a consistent decline in the dominant frequency of EEG activity from the stimulus until the seizure endpoint, a phenomenon which has now been quantitatively verified (Zoldi et al., 1996, Krystal et al., In Press, West et al., In Press). The typical spatial properties of EEG data recorded during ECT seizures have been investigated using multi-channel EEG recordings. Evidence suggests that the amplitude of ictal EEG activity varies significantly across the scalp, with maximum amplitude typically manifest in fronto-central sagital areas (Enderle et al., 1986, Weiner et al., 1991) while the lowest amplitude tends to be in the temporal, pre-frontal, and occipital regions. These latter locations also reflect the greatest amplitude variation among seizures (greatest coefficient of variation of the amplitude) (Krystal, Greenside, et al., 1996). In terms of ictal EEG amplitude, UL ECT has been associated with a tendency for greater amplitude in the stimulated hemisphere while BL ECT seizures are more symmetrical (Staton et al., 1981, Brumback and Staton, 1982, Enderle et al., 1986, Weiner et al., 1991, Weiner and Krystal, 1993). Variation among seizures is also reported in the degree of correlation of EEG activity between regions. Preliminary analysis suggests that EEG data recorded from leads closest to the vertex have the greatest degree of correlation, whereas the prefrontal ictal EEG data is less well correlated with data recorded from other brain regions (Zoldi et al., 1996, Krystal, Zoldi et al., In Press). In terms of regional inter-relationships of EEG activity, there is also evidence that in some seizures, there appears to be a significant temporal delay in the manifestation of ictal EEG changes in certain recording sites, most notably temporal, prefrontal, and occipital areas (Zoldi et al., 1996, Krystal et al., In Press). In addition, we have found evidence for a consistent anterior-to-posterior time delay in mid-ictal EEG rhythmic activity, and that this delay is suggestive of electrical wave behavior (Zoldi et al., 1996, Prado, 1998, Krystal et al., In Press). These phenomena indicate the presence of complex and highly variable spatio-temporal behavior in ictal EEG data. In the section below we discuss what is known about how this behavior relates to ECT outcome. B. Relationships Between the Ictal EEG and Treatment Outcome The variability of the ictal EEG manifestations of ECT seizures is consistent with recent evidence that contradicts the long held view that ECT seizures are all-or-none phenomena (Krystal and Weiner, 1994, Krystal, 1998). These studies suggest that significant outcome differences occur with different types of ECT treatment electrical stimuli; stimuli higher above seizure threshold (i.e. higher relative stimulus intensity) are associated with greater therapeutic efficacy and cognitive side-effects than are barely suprathreshold stimuli (Sackeim et al., 1987, 1991, 1993, Krystal et al., 1995, 1998, Krystal, Weiner, and Coffey, In Press). This difference in therapeutic efficacy is particularly notable for unilateral (UL) ECT where evidence suggests that antidepressant effects of barely supratheshold treatments may be no better than placebo (Sackeim et al., 1987, 1991, 1993), while for BL ECT higher intensity stimuli appear to only increase the speed of response. Recent evidence also suggests that these outcome differences among types of ECT are paralleled by physiologic differences in the EEG. Ictal EEG data recorded during the more efficacious, higher intensity stimuli differ in a number of respects from the seizures elicited by the less effective barely suprathreshold UL treatments including: a) higher amplitude of 2-5 and 5-13 Hz activity immediately after the stimulus, b) greater immediate post-stimulus interhemispheric EEG coherence (correlation of low-frequency EEG activity between the two hemispheres), c) a shorter time between the electrical stimulus and the onset of high amplitude slow- waves (less than 5 Hz), d) increased amplitude of 2-5 and 13-30 Hz activity in the mid-ictal portion of the seizure, e) more regular mid-ictal slow-wave shape (or signal predictability over time), f) more pronounced postictal suppression, g) less postictal interhemispheric coherence, and h) a greater and more rapid decline in the dominant frequency of EEG activity over the seizure (Nobler et al., 1993, Krystal et al., 1993, 1995, 1996, 1998, Krystal, 1998, Krystal and Weiner, In Press, West et al., In Press). These studies suggest an association between more efficacious types of ECT treatment and ictal EEG features, thereby raising the possibility that there may be features of the ictal EEG that directly reflect treatment outcome. A number of studies have in fact confirmed that such a relationship exists. Five studies, employing qualitative ratings of ictal EEG activity, have indicated that greater post-ictal suppression, higher ictal amplitude, and greater ictal morphologic regularity are associated with a better therapeutic outcome (Nobler et al., 1993, Hrdlicka et al., 1996, Folkerts, 1996, Suppes et al., 1996, Krystal et al. 1998). In addition, our group has carried out studies involving quantitative computer analysis of ictal EEG data which suggest that the following features are predictive of better therapeutic outcome: 1) greater ictal amplitude, 2) shorter time to the onset of high-amplitude slow-waves, 3) greater ictal coherence, 4) greater morphologic regularity, 5) lower post-ictal amplitude, and 6) lower post-ictal coherence (Krystal et al., 1995, Krystal et al., 1996, Krystal et al., 1998, Krystal, 1998, Krystal and Weiner, In Press). There is also the issue of whether peri-ictal EEG measures relate to adverse effects of ECT. In this regard we carried out a preliminary UL ECT study which demonstrated a relationship between quantitative features of the ictal EEG and some objectively assessed effects. Specifically, greater ictal EEG amplitude and postictal suppression were both associated with diminished performance on complex figural memory and recent autobiographical memory recall tasks. These relationships are notable for the fact that they persisted after the effects of relative stimulus intensity had been taken into account. C. Implications for ECT The studies described above indicate that more effective ECT treatments appear to be associated with seizures that (1) manifest earlier, more morphologically predictable, and more intense slow-wave EEG activity; (2) maintain a higher level of ictal interhemispheric correlation; and (3) display greater immediate postictal amplitude suppression, along with diminished interhemispheric correlation during the postictal period. Preliminary evidence also suggests that some of these features may be associated with more severe cognitive side-effects. The existence of these relationships has led us to investigate the potential utility for ictal EEG measures to aide in the actual clinical dosing of ECT treatments (Krystal and Weiner, 1994, Krystal, 1998). In this regard, EEG features might be expected to allow clinicians to better predict the degree of therapeutic response and cognitive side-effects associated with ECT, something that would otherwise not be apparent clinically until an appreciable period of time, sometimes up to several weeks, had elapsed (Krystal and Weiner, 1994, Krystal, 1998). On the basis of such measurements, clinicians might be better able to maximize the therapeutic effects of treatment, while minimizing the cognitive side-effects, than allowable by present present dosing strategies (Krystal and Weiner, 1994, Krystal et al., 1998, Krystal, 1998). Alternatively, since it has not been established whether a minimum threshold of physiologic impact is necessary for a therapeutic response, such models may be useful for allowing practitioners to maintain a stimulus intensity level over the entire treatment course that is able to achieve a predictable tradeoff in efficacy and side-effects, thereby optimizing treatment delivery (Krystal, 1998). Recent studies confirm that ictal EEG models are likely to allow clinicians improved ability to maintain a desired relative stimulus intensity over the ECT treatment course (Krystal et al., 1995, 1998, Krystal, 1998). However, while such preliminary data demonstrates that a multi-variate ictal EEG model of treatment outcome is able to differentiate therapeutic responders and non-responders, further investigation is needed to determine whether dosing of ECT treatments on the basis of such models actually improves ECT outcome (Krystal, 1998). Physiologically, the available data on relationships between ictal EEG data and ECT treatment outcome are all suggestive that seizures having a more intense and more widespread physiologic effect on brain function are associated with a better therapeutic response but also greater cognitive side-effects. However, since this data has primarily focused on prefrontally-derived EEG data, the extent to which other brain regions are involved remains unknown. The available studies also do not indicate a laterality effect in that left and right prefrontally-derived EEG data are both predictive of outcome. Further research is needed to sort out these issues. EFFECTS OF ECT ON THE INTER-ICTAL EEG A. Properties of the Interictal EEG with ECT The most pronounced change in the EEG over a course of ECT is an increase in the amount of slow activity (activity that is less than 8.5 Hz in the waking EEG) (Fink, 1979, Weiner, 1983, Krystal and Weiner, In Press). This slowing, which includes both rhythmic and arrhythmic components, builds up over the treatment course and begins to diminish immediately after the last treatment. It is largely gone by 1 month following the ECT course, and is only rarely present past 3 months (Weiner, 1980,1983, Weiner et al., 1986a). The topological distribution of this activity is diffuse and frontally predominant (Weiner et al., 1986a, Rosen and Silfverskold, 1987). The increase in slowing in the EEG is similar to the change which takes place in the waking EEG in individuals experiencing a wide variety of generalized encephalopathic states, including spontaneous seizures, and has therefore traditionally been viewed as nonspecific (Weiner, 1983, Krystal and Weiner, In Press). A decrease in the average EEG background frequency and in the amount of higher frequency beta activity have also been reported to occur with ECT, however these effects are of uncertain clinical significance (Fink, 1979, Weiner, 1983). Activity of a paroxysmal or epileptiform nature is not a typical effect of ECT (Weiner, 1983). A number of factors have been reported to affect EEG slowing associated with ECT. The severity and persistence of EEG slowing has been reported to be correlated with the number of ECT treatments administered during the treatment course (Pacella et al., 1942, Mosovich and Katzenelbogen, 1948, Roth, 1951). In addition, an abnormal EEG prior to the ECT course increases the likelihood of greater slowing and appears to delay the time until this slowing resolves following the end of the ECT course (Turek, 1972, Weiner, 1983). A greater amount of EEG slowing has also been reported with BL as compared with UL ECT (Small et al., 1970, 1973, Abrams et al., 1972, Stromgren and Juul-Jensen, 1975, Weiner, 1983). Further, there is at times a greater amount of slowing induced in the stimulated hemisphere with UL ECT (d'Elia and Perris, 1970, Abrams et al., 1970, Volavka, 1972, Stromgren and Juul-Jensen, 1975, Weiner, 1983). Finally, brief-pulse stimuli induce less slowing than sine-wave ECT stimuli (Weiner, 1983, Weiner et al., 1986a). One study examining the effects of ECT on the coherence of activity among the scalp leads in EEG reported that lower frequency (less than 8.5 Hz) intra- and inter-hemispheric coherence is increased over the ECT course, while the coherence in higher frequencies tends to be disrupted (Krystal et al., 1991). In particular, ECT appears to acutely increase the low-frequency interhemispheric and intrahemispheric coherence in fronto- central regions. In contrast, ECT produces decreased 8.5-13 Hz coherence between the hemispheres posteriorly, while within hemisphere coherence is diminished in frontal, central, and temporal regions. These coherence effects are no longer detectable one month after the end of the treatment course. The coincidence of an increase in fronto-central inter- and intra-hemispheric slow-wave coherence, along with a frontally predominant slow- wave amplitude increase, suggests that the slowing induced by ECT is largely correlated among the involved leads. B. Relationships between the background EEG and the therapeutic effect of ECT Whether there is a relationship between background EEG activity and ECT therapeutic response has remained uncertain (Weiner, 1983, Sackeim et al., 1996, Krystal and Weiner, In Press). Several studies have reported that greater EEG slowing was associated with a better therapeutic response (Proctor and Goodwin, 1943, Roth, 1951, 1952, Roth et al., 1957, Fink and Kahn, 1957, Ottosson, 1962, Stromgren and Juul-Jensen, 1975, Abrams et al., 1987, Sackeim et al., 1996), while some studies reported poorer outcome with greater slowing (Honcke and Zahle, 1946, Mosovich and Katzenelbogen, 1948), and still others did not find any relationship between EEG slowing and therapeutic response (Taylor and Pacella, 1948, Chusid and Pacella, 1952, Bergman et al., 1952, Johnson et al., 1960, Ulett, 1962, Sutherland et al., 1969, Abrams et al., 1970, 1972, Volavka et al., 1972, Volavka, 1974, Kurland et al., 1976, Weiner et al., 1986a, Drake and Shy, 1989, Abrams et al., 1992, Malaspina et al., 1994). Variability in the methodology of these studies appears to have contributed to the disagreement among them. One such methodologic factor is that most studies investigating the relationship between EEG slowing and therapeutic response employed unstandardized qualitative visual rating of EEG data. While investigations utilizing quantitative EEG analysis can be much more easily standardized, such studies have varied in other methodologic factors such as: the use of different numbers and placement of recording electrodes, the definitions of EEG frequency bands, the degree of care employed in excluding artefact-contaminated data, and the rigor of the statistical analytic techniques (Krystal and Weiner, In Press). One recent methodologically rigorous study reported a relationship between the therapeutic response to ECT and changes in the inter-ictal EEG (Sackeim et al., 1996). These investigators carried out quantitative analysis of 19 channel background EEG data that had been carefully screened for artifacts. They compared background EEG data collected prior to and after a course of ECT in 62 patients with major depression who had been screened using standardized diagnostic criteria. A data analytic technique referred to as the "Scaled Subprofile Model" was employed to allow analysis of spatial patterns in EEG data recorded from multiple leads over a number of frequency bands, while correcting for the tendency to incorrectly find a significant difference by chance when making multiple comparisons (minimizing Type I errors) (Sackeim et al., 1996, Krystal and Weiner, In Press). After employing these careful methodologic techniques, Sackeim and colleagues found that greater therapeutic response was associated with a greater increase in frontal and pre-frontal low-frequency (0.5- 3.5 Hz) spectral amplitude over the treatment course (Sackeim et al., 1996). Using similar methodology we have since replicated this finding (Krystal et al., 1997). C. Relationships between the background EEG and the cognitive side-effects of ECT Given the well known relationship between the presence of slow activity in the waking EEG and encephalopathy, as well as reports of greater EEG slowing from types of ECT treatment associated with greater cognitive impairment (sine-wave and BL ECT), it is surprising that relatively few studies have examined the relationship between interictal EEG slowing and the cognitive side-effects of ECT (Weiner, 1983, Krystal and Weiner, In Press). Several studies suggest that greater EEG slowing appears to be associated with a greater cognitive side-effects with ECT. Earlier studies reported that greater EEG slowing was associated with greater memory loss with ECT (Fink et al., 1961, Stromgren and Juul-Jensen, 1975, Weiner et al., 1986b). More recently, a quantitative correlation was reported between an increase in EEG slowing (0.5-4 Hz) specific to the left temporal and fronto-central regions and the degree of effect on delayed verbal paired associate recall with ECT (Krystal et al., 1991). This study also found that a greater increase in 0.5-4 Hz coherence between the fronto-central regions in the two hemispheres was also associated with a greater effect on delayed verbal paired associate recall (Krystal et al., 1991). D. Implications for ECT The relationship between the inter-ictal EEG and the response to ECT remains somewhat uncertain. The results of studies on the relationship to therapeutic response has been contradictory and relatively few studies have been carried out on the relationship between the inter-ictal EEG and the memory effects of ECT. A few recent studies employing quantitative EEG analysis and utilizing relatively rigorous methodology suggest relationships between inter-ictal EEG changes and both therapeutic response and cognitive side-effects. Greater slowing in frontal and prefrontal regions appears to be associated with greater therapeutic response (Sackeim et al., 1996, Krystal et al., 1997). In turn, greater slowing in the left temporal and fronto-central regions and correlation of low frequency activity between the left and right fronto-central regions are associated with greater verbal memory effect (Krystal et al., 1991). Given that both therapeutic response and cognitive side-effects are associated with increased frontal slowing one would expect there to be a robust relationship between efficacy and memory outcome measures. However, a considerable body of data indicate that the therapeutic effects of ECT are independent of the induced cognitive side- effects (Weiner, 1983, Weiner et al., 1986b). This means that the relationships between inter-ictal EEG indices such as slowing and ECT outcome measures are complex, either with respect to the amplitude of the induced slowing or with regard to its topographic distribution. One possible explanation is that the presence of a small to moderate degree of slowing represents a physiologic index of treatment adequacy while totally absent slowing reflects ineffective treatments, and large amounts of slowing is associated with a high likelihood of memory impairment without further enhancement of efficacy. Alternatively, memory effects may have a different topographical relationship to EEG slowing than does efficacy. For example, prefrontal slowing could be nonspecific in nature or related to treatment efficacy, while temporal slowing, particularly in the dominant cerebral hemisphere, could be specific to memory effects. The fact that unilateral non-dominant ECT is associated with less cognitive impairment than BL ECT, even when effective, and is accompanied by less slowing in general but particularly in the dominant hemisphere (d'Elia and Perris, 1970, Abrams et al., 1970, Small et al., 1970, 1973, Abrams et al., 1972, Volavka, 1972, Stromgren and Juul-Jensen, 1975, Weiner, 1983), suggests that such a topographic model may have value. Further studies are needed to determine the specificity of the relationships between regional slowing and ECT outcome and thereby test this hypothesis. If specific and independent relationships are found they would also be of great importance because of their potential to steer the development of new types of ECT to give practitioners improved ability to maximize therapeutic response while minimizing cognitive side-effects. IMPLICATIONS FOR rTMS Because both ECT and rTMS involve electromagnetic stimulation of the brain and are of interest as antidepressant treatments, we will next consider what implications the above EEG effects of ECT have for rTMS. The use of rTMS as a treatment for major depression and it's effects on cerebral neurophysiology are reviewed elsewhere in this issue, as are the differences between electrical and magnetic stimulation (see Boutros et al., Pascual Leone et al., Pridmore et al., Szuba et al., This Issue). A fundamental difference between the present use of rTMS and ECT as potential treatments for major depression is that while the former involves subconvulsive stimulation, evidence suggests that inducing a seizure is central to efficacious ECT (Sackeim, 1994a, George and Wasserman, 1994). The mechanisms of the therapeutic effect of subconvulsive rTMS, however, remain unknown (see Lisanby and Belmaker, Pascual Leone et al., Szuba et al., Pridmore et al., This Issue). Thus, the EEG effects of seizures elicited with ECT may have no bearing whatsoever on rTMS. On the other hand, some neurophysiologic changes associated with rTMS are reminiscent of changes that occur with ECT that appear to be linked to the EEG findings described above and which may have some relationship to the mechanism of action of ECT. A number of studies suggest that ECT and both electroconvulsive shock and rTMS in animals, cause an increase in the seizure threshold (Sackeim et al., 1986, 1987, 1991, 1993, Krystal et al., 1998, Sackeim, In Press). One important hypothesis regarding the therapeutic mechanism of action of ECT is that this increase in seizure threshold is related to the ability of the induced seizures to elicit active anticonvulsant processes which mediate the antidepressant effects of treatment (Sackeim et al., 1983, Post et al., 1986, Post, 1990, Sackeim, 1994b, In Press). Some evidence suggests that the slow-waves elicited during ECT seizures reflect the evocation of these inhibitory processes which then become manifest in immediate postictal suppression, and subsequent inter-ictal EEG slowing (Margerison and Corsellis, Sackeim et al., 1986, 1991, Krystal et al., 1993, 1998, Nobler et al., 1993, Krystal, 1998, Sackeim, In Press, Krystal and Weiner, In Press). The data on rTMS in rats, indicating an increase in the seizure threshold to electroconvulsive shock, suggests that such a change might also be expected to accompany rTMS in humans (Fleischmann et al., 1995). It is particularly notable that not only did rTMS cause an increase in seizure threshold, but it was also accompanied by the same changes in behavioral models of depression which also occur with maximal electroshock. Thus, an increase in anticonvulsant effects accompanying rTMS may relate to it's therapeutic antidepressant effect and may point to a common pathway of mechanism of action between ECT and rTMS and possible relevance for rTMS of the ictal and inter-ictal EEG findings with ECT described above. In this regard, one issue of potential physiologic relevance is the determination of what brain regions should be stimulated with rTMS. While ECT involves a relatively global distribution of current throughout the brain (Sackeim, 1994), EEG correlations with outcome are much more spatially localized. Available ictal EEG studies with ECT suggest an importance of prefrontal activation, though they have generally only included left and right prefrontal (referenced to ear or mastoid) EEG leads. Thus, the ictal EEG data do not allow a determination of the relative utility of stimulation in other regions. In addition, such investigations do not differentiate between left and right prefrontal regions in terms of physiologic physiologic linkage to therapeutic antidepressant response. In contrast, available inter-ictal EEG data, which are based on more spatially comprehensive recordings, indicate a specificity for frontal and prefrontal regions with respect to correlations with therapeutic outcome, again without lateralization of effect. However, both left temporal and fronto-central effects appear to be associated with an increase in cognitive side-effects. The data suggests that left prefrontal, right prefrontal, or right frontal rTMS might be most likely to be associated with therapeutic benefit with least verbal memory impairment. Present data on this issue indicates that prefrontal rTMS does appear to produce an antidepressant effect, however this may be limited to only left hemispheric stimulation (Pascual-Leone et al., 1996, George et al., 1997). There are two other areas in which ECT-related neurophysiologic observations may have relevance for rTMS. These are the potential of rTMS to elicit focal seizures and the theoretical use of rTMS to inhibit certain regions of the brain during generalized tonic-clonic seizures induced by ECT and therefore focally disrupt such activity (Sackeim, 1994a, George and Wassmerman, 1994). In this regard it is interesting to note that this same idea of eliciting partial seizures to elicit a therapeutic psychiatric effect with less side-effects than traditional ECT emerged very early in the history of ECT (Impastato and Pacella, 1952, Impastato et al., 1953). These investigators elicited focal seizures by placing treatment electrodes across the motor strip and gradually increasing the stimulus current. Initially they observed contractions of the contralateral limb and arm in synchrony with stimulus pulses. When approximately 10 milliAmperes (mA) was exceeded they tended to see the onset of contralateral tonic motor activity. Between 10-20 mA, a unilateral tonic convulsive response was typically manifest throughout the contralateral side of the body and apnea ensued. Once this was achieved the current was shut off or reduced to 2-5 mA. The unilateral tonic convulsion typically lasted 5-15 seconds and was followed by a 5-10 second unilateral clonic convulsive phase. If any movement occurred on the ipsilateral side, the electrodes were moved closer together, or slightly more posteriorly or laterally. In this manner Impastato and coworkers were able to produce seizures with electrical stimulation that were focal. Unfortunately, there are not any blinded and controlled studies on the efficacy and side-effects associated with this form of treatment. There is, however, one qualitative description of therapeutic outcome with this electrically induced focal seizure technique (Impastato et al., 1953), which reported that there some patients did not show an adequate response to focal stimulation treatment but who did respond to traditional ECT. As a result, these investigators abandoned this procedure for a related technique in which the initiation of a focal seizure was followed by an increase in stimulus current until a bilateral convulsion was achieved (Impastato et al., 1953). The same group of investigators studied the inter-ictal EEG effects associated with the focal seizure technique, reporting that "the records were characterized by bilateral slow activity of the same type as that seen after grand mal seizures..... In general, the changes were less marked after focal seizures and they occurred in a smaller proportion of patients." (Bergman et al., 1953). The ability to elicit focal seizures with electrical stimulation is of relevance to the potential for rTMS to do the same, particular given the increased focality of the rTMS (Sackeim, 1994, George and Wasserman, 1994). As Sackeim has pointed out, the issue may well not be the intensity or generalization of seizure activity but the ability of the seizure to affect the physiology of particular structures (Sackeim, 1994a). While the efficacy associated with electrical focal seizure induction may have been limited (though not systematically studied), the inter-ictal ECT literature suggests that rTMS elicited focal seizures in other regions, in particularly, prefrontally, associated with focal prefrontal inter-ictal slowing may be more therapeutic. Still, there is no evidence that focal seizures can be reliably elicited with rTMS and there are a number of technical problems that would have to be overcome (Sackeim, 1994a). For that matter, the only seizures so far induced with rTMS have been generalized (Wasserman, 1998). Risk of seizure generalization is also a concern given findings in the epilepsy literature that frontal and pre-frontal epileptogenic foci often precipitate generalized tonic-clonic seizures (Niedermeyer, 1993). Such seizures would be dangerous outside the setting of muscular relaxation and oxygenation used with ECT and would eliminate the potential advantage of focality in terms of cognitive side-effects. Clearly, extensive experimental investigation in animal models of the feasibility and safety of focal seizure induction with rTMS will be necessary before human studies can be considered. Perhaps a more feasible goal may be to use rTMS to inhibit and therefore protect specific regions of interest such as left anterior-temporal, and fronto-central regions during generalized tonic-clonic seizures produced with either electrical or magnetic stimulation (Sackeim, 1994a). Theoretically, this might be expected to result in diminished cognitive side-effects, without compromising therapeutic outcome. However, it remains to be determined whether a generalized tonic-clonic seizure can be initiated and maintained while regions are being stimulated with rTMS, whether such stimulation is able to suppress the development of seizure activity in underlying neuronal arrays, and whether such inhibition prevents the developments of side-effects while allowing therapeutic changes to occur. These are all questions that can and should be addressed in future experimental and clinical studies. SUMMARY AND CONCLUSIONS Recent studies indicate that there are consistent relationships between the ictal and inter-ictal EEG effects of ECT and the response to treatment. Earlier and more intense onset of slow-waves in the prefrontal ictal EEG with better correlation between the hemispheres is associated with a greater therapeutic response. Greater prefrontal EEG amplitude suppression in the immediate post-ictal period is also associated with a more robust antidepressant response. At the same time, the ability of a course of ECT to elicit pre-frontal slowing in the inter-ictal EEG appears to be related to the antidepressant efficacy of the treatments, while greater left anterior temporal slowing and coherence of slow activity in left temporal and fronto-central regions is associated with greater verbal memory impairment. While further work is needed to determine the regional specificity of these findings (particularly the ictal EEG findings), they are consistent with recent rTMS work indicating the antidepressant potential of prefrontal rTMS. The possibility of eliciting focal frontal seizure activity with rTMS or using rTMS to inhibit seizure activity in left temporal and frontocentral regions might also be expected to have antidepressant efficacy with less side-effects than ECT on the basis of this work, however, extensive experimentation will be needed before the feasibility of these approaches can be reasonably assessed. ACKNOWLEDGEMENTS The authors would like to acknowledge the support of a NARSAD Young Investigator Award and NIMH Grants K20MH01141 and R29MH57532. Also, Dr. Krystal and Dr. Weiner are inventors on a patent related to the use of ictal EEG data as an indicator of ECT seizure adequacy which is issued to Duke University and licensed to MECTA Corporation, Lake Oswego, Or. All inventor royalties deriving from this patent are being donated to a charitable organization. REFERENCES American Psychiatric Association (APA). The practice of ECT: Recommendations for treatment, training, and privileging. A task force report of the American Psychiatric Association. Washington, DC: American Psychiatric Press. 1990. Abrams R, Fink M, Feldstein S, Volavka J, Roubicek J (1972). Unilateral and bilateral electroconvulsive therapy: effects on depression, memory and the electroencephalogram. Arch. Gen. Psychiatry. 27:88-94. Abrams R, Volavka J, Roubicek J, Dornbush R, Fink M (1970). Lateralized EEG changes after unilateral and bilateral electroconvulsive therapy. Dis. Nerv. Syst. 31:28-33. Bergman Ps, Impastato DJ, Berg S, Feinstein R (1953). Electroencephalographic changes following electrically induced focal seizures. Confin. Neurol. 13:271-277. Brumback RA, Staton RD (1982). The electroencephalographic pattern during electroconvulsive therapy. Clin Electroenceph. 13:148-153. D'Elia g, Perris C (1970). Seizure and post-seizure electroencephalographic pattern. Acta Psychiatr. Scand. (Suppl) 215:9-29. Enderle JD, Staton RD, Gerst JW, Barr CE, Brumback RA (1986). The electroencephalographic pattern during electroconvulsive therapy III. Analysis of frontotemporal and nasopharyngeal spectral energy. Clin. Electroenceph. 17:66-77. Fink M (1979). Convulsive Therapy: Theory and Practice. New York: Raven Press. Fink M, Kahn RL, Karp AE, Pollack A, Green NA, Lefkowitz AB (1961). Inhalant induced convulsions: significance for the theory of the convulsion therapy process. Arch Gen Psychiatry. 4:259-66. Fleischmann A, Prolov K, Abarbanel J, Belmaker RH (1995). The effect of transcranial magnetic stimulation of rat brain on behavioral models of depression. Brain Research. 699:130-132. Folkerts H. (1996).The ictal electroencephalogram as a marker for the efficacy of electroconvulsive therapy. Eur. Arch. Psychiatry Clin. Neurosci. 246:155-164. Fleming GWTH, Golla FL, Walter WG. (1939). Electroconvulsive therapy of schizophrenia. Lancet. 2:1353- 1355. George MS, Wasserman EM (1994). Rapid-rate transcranial magnetic stimulation and ETC. Convulsive Therapy. 10:251-254. George MS, Wasserman EM, Kimbrelll TA, Little JT, Williams WE, Danielson AL, Greenberg BD, Hallet M, Post RM (1997). Mood improvement following daily left prefrontal repetitive transcranial magnetic stimulation in patients with depression: A placebo controlled crossover trial. Am. J. Psychiatry. 154:1752-1756. Honcke P, Zahle V. (1946). On the correlation between clinical and electroencephalographic observations in patients treated with electroshock. Acta Psychiatr. Neurol. Scand. (Suppl.) 47:451-58. Hrdlicka M, Moran M, Vachutka J, Blatny M, Rozinkova J (1996). EEG in electroconvulsive therapy. Is more intensive paroxysmal activity associated with a higher therapeutic response? Neuropsychobiology. 33:138-141. Impastato DJ, Pacella BL (1952). Electrically produced unilateral convulsions: A new method of electroconvulsive therapy. Diseases of the nervous system. 13:368-369. Impastato DJ, Berg S, Pacella BL (1953). Electroshock therapy: Focal spread technic. A new form of treatment of psychiatric illness. Conf. Neurol. 13:266-271. Izumi S, Takase M, Arita M, Masakado Y, Kimura A, Chino N (1997). Transcranial magnetic stimulation- inducedc changes in EEG and responses recorded from the scalp of healthy humans. Electroenceph. Clin. Neurophysiol. 103:319-322. Krystal AD (1998). The clinical utility of ictal EEG seizure adequacy models. Psychiatric Annals. 28:30-35. Krystal AD, Coffey CE, Weiner RD, Holsinger T. (1998). Changes in seizure threshold over the ECT course affect therapeutic response and are detected by ictal EEG ratings. Jl. Neuropsychiatry Clin Neurosci. 10:178- 186. Krystal AD, Coffey CE, Weiner RD, Rapp PE, Albano A, Greenside HS, DeMasi M, Cellucci C. (1997). EEG correlates of the response to ECT. Biol. Psychiatry. 41:56S. Krystal AD, Greenside HS, Weiner RD, Gassert D (1996). A comparison of EEG signal dynamics in the waking EEG, after anesthesia induction, and during ECT seizures. Electroenceph. Clin. Neurophysiol.. 99:129-140. Krystal AD, Zoldi SM, Prado R, Greenside HS, West M (In Press). The spatiotemporal dynamics of generalized tonic-clonic seizure EEG data: Relevance to the clinical practice of electroconvulsive therapy. In: Non-linear Dynamics in Brain Function. Krystal AD, Weiner RD (1994). ECT seizure therapeutic adequacy. Convulsive Therapy. 10:153-164. Krystal AD, Weiner RD (In press). EEG correlates of the response to ECT: A possible antidepressant role of brain-derived neurotrophic factor. J. of ECT. Krystal AD, Weiner RD, Coffey CE (In Press). The relationship between relative stimulus intensity, the ictal EEG, and unilateral ECT cognitive side-effects. J. of ECT. Krystal AD, Weiner RD, Coffey CE (1995). The ictal EEG as a marker of adequate stimulus intensity with unilateral ECT. J. Neuropsychiatry Clin. Neurosci. 7:d295-303 Krystal AD, Weiner RD, Coffey CE, Smith P, Arias R (1991). EEG coherence before and after ECT Proceedings of the American Psychiatric Association Meeting. Krystal AD, Weiner RD, Gassert D, McCall WV, Coffey CE, Sibert T, Holsinger T. (1996). The relative ability of three ictal EEG frequency bands to differentiate ECT seizures on the basis of electrode placement, stimulus intensity, and therapeutic response. Convulsive Therapy. 12:13-24. Krystal AD, Weiner RD, McCall WV, Shelp FE, Arias R, Smith P (1993). The effects of ECT stimulus dose and electrode placement on the ictal electroencephalogram: An intra-individual cross-over study. Biological Psychiatry. 34:759-767. Krystal AD, Zaidman C, Greenside HS, Weiner RD, Coffey CE. (1997) The largest Lyapunov exponent of the EEG during ECT seizures as a measure of ECT seizure adequacy. Electroenceph. Clin. Neurophysiol. 103:599- 606. Margerison JH, Corsellis JA (1966). Epilepsy and the temporal lobes: A clinical, electroencephalographic, and neuropathological study of the brain in epilepsy, with particular reference to the temporal lobes. Brain. 89:499- 530. Mosovich A, Katzenelbogen S (1948). Electroshock therapy. Clinical and EEG studies. J. Nerv. Ment. Dis. 107:517-530. Niedermeyer E (1993) Epileptic Seizure Disorders. In: Electroencephalography: Basic Principles, Clinical Applications and Related Fields. Eds. Niedermeyer E, Lopes Da Silva F. Williams and Wilkins. Baltimore MD. pp 461-564. Nobler MS, Sackeim HA, Solomou M, Luber B, Devanand DP, Prudic J (1993). EEG manifestations during ECT: Effects of electrode placement and stimulus intensity. Biological Psychiatry. 34:321-330. Pacella BL, Barrera SE, Kalinowsky LB (1942). Variations in the EEG with EST of patients with mental disorders. Arch. Neurol. Psychiatry 47:367-84. Pascual-Leone A, Pallardo RB, Catala D (1996). Beneficial effect of rapid-rate transcranial magnetic stimulation of the left dorsolateral prefrontal cortex in drug-resistant depression. Lancet. 348:233-238. Post RM (1990). ECT: The anticonvulsant connection. Neuropyschopharm. 3:89-92. Post RM, Putnam F, Uhde TW, Weiss SR (1986). Electroconvulsive therapy as an anticonvulsant. Implications for its mechanism of action in affective illness. Ann. N. Y. Acad. Sci. 462:376-388. Prado R (1998). Latent structure in non-stationary time series. Ph.D. Thesis, Institute for Statistics and Decision Sciences, Duke University. Proctor LD. Goodwin JE. Comparative EEG observations following ECT using raw 60 cycle alternating and unidirectional fluctuating current. (1943). Am. J. Psychiatry. 99:525-530. Roemer RA, Shagass C, Dubin W, Jaffe R, Katz R (1991). Relationship between pretreatment electroencephalographic coherence measures and subsequent response to electroconvulsive therapy: A preliminary study. Neuropsychobiology. 24:121-124. Rosen I, Silfverskold P (1987). Qunatification of EEG changes following electroconvulsive therapy in Depression. Eur Arch. Psychiatr Neurol Sci. 236:209-213. Roth M (1951). Changes in the EEG under barbiturate anaesthesia produced by ECT and their significance for the theory of ECT action. Electroenceph. Clin. Neurophysiol. 30:62-72. Sackeim HA (1994a). Magnetic stimulation therapy and ECT. Convulsive Therapy. 10:255-258. Sackeim HA (1994b). Central issues regarding the mechanisms of action of ECT: Directions for future research. Psychopharm. Bull. 30:281-308. Sackeim HA (In Press). The anticonvulsant hypothesis of the mechanisms of action of ECT: Current Status. Sackeim HA, Decina, Prohovnik I, Malitz S, Resor SR (1983). Anticonvulsant and antidepressant properties of ECT: A proposed mechanism of action. Biol. Psychiatry. 18:1301-10. Sackeim HA, Decina P, Prohovnik I, Portnoy S, Kanzler M, Malitz S (1986). Dosage, seizure threshold, and the antidepressant efficacy of electroconvulsive therapy. Ann. N. Y. Acad. Sci. 462:398-410. Sackeim HA, Decina P, Prohovnik I, Malitz S (1987). Seizure threshold in electroconvulsive therapy. Effects of sex, age, electrode placement and number of treatments. Arch. Gen. Psychiatry. 44:355-60. Sackeim HA, Devanand DP, Prudic J (1991). Stimulus intensity, seizure threshold, and seizure duration: Impact on the efficacy and safety of electroconvulsive therapy. Psych. Clin. North Am. 14:803-43. Sackeim HA, Prudic J, Devanand DP, Kiersky JE, Fitzsimons L, Moody BJ, McElhiney MC, Coleman EA, Settembrino JM (1993). Effects of stimulus intensity and electrode placement on the efficacy and cognitive effects of ETC. N. Engl J. Med. 328:839-46. Sackeim HA. Luber B, Katzman GP, Moeller JR, Prudic J, Devanand DP, Nobler MS. (1996). The effects of electroconvulsive therapy on quantitative electroencephalograms. Arch. Gen. Psychiatry. 53:814-824. Small IF, Small JG, Milstein V, Sharpley P (1973). Interhemispheric relationships with somatic therapy. Dis. Nerv. Syst. 34:170-177. Small JG, small IF, Perez HC, Sharpley P (1970). Electroencephalographic and neurophysiological studies of electrically induced seizures. J. Nerv. Ment. Dis. 150:479-489. Staton RD, Hass PJ, Brumback RA (1981). Electroencephalographic recording during bitemporal and unilateral non-dominant hemisphere (lancaster position) electroconvulsive therapy. J. Clin Psychiatry. 42:264-269. Stromgren LS, Juul-Jensen P (1975). EEG in unilateral and bilateral electroconvulsive therapy. Acta Psychiatr. Scand. 51:340-360. Suppes T, Webb A, Carmody T, Gordon E, Gutierrez-Esteinou R, Hudson JI, Pope HG. (1996). Is postictal electrical silence a predictor of response to electroconvulsive therapy? J Affective Disorders. 41:55-58. Turek IS (1972). EEG correlates of electroconvulsive treatment. Dis. Nerv. Syst. 33:584-89. Volavka J (1972). Neurophysiology of electroconvulsive therapy. Semin. Psychiatry. 4:55-65. Wassermann EM (1998). Risk and safety of repetitive transcranial magnetic stimulation: report and suggested guidelines from the international workshop on the safety of repetitive transcranial magnetic stimulation, June 5- 7, 1996. Electroenceph. Clin Neurophysiol. 108:1-16. Weiner RD. (1983) EEG related to electroconvulsive therapy. In: EEG and Evoked Potentials in Psychiatry and Behavioral Neurology. Edited by: Hughes JR, Wilson WP. Butterworth Publishers. Boston. pp. 101-126. Weiner RD, Rogers HJ, Davidson JRT, Kahn EM (1986a). Effects of electroconvulsive therapy upon brain electrical activity. Annals New York Academy of Sci. 462:270-281. Weiner RD, Rogers HJ, Davidson JRT, Squire LR (1986b). Effect of stimulus parameters on cognitive side- effects. Annals New York Academy of Sci. 462:315-325. West M, Prado R, Krystal AD (In Press). Evaluation and comparison of EEG traces: Latent structure in non- stationary time series. Jl. Of the American Statistical Association. Zoldi S, Krystal AD, Greenside HS (1996). Redundancy and stationarity of 21-channel EEG data recorded during ECT seizures. 13:440.