Electroconvulsive Therapy

Harold A. Sackeim, D. P. Devanand, and Mitchell S. Nobler

Convulsive therapy was introduced in 1934 by Meduna. He produced chemically induced generalized seizures in schizophrenic patients, based on the mistaken belief that schizophrenia and epilepsy were mutually incompatible (1). Despite a false rationale, it was soon evident that convulsive therapy often resulted in dramatic clinical improvement in psychiatric patients. At the time of the near simultaneous introduction of convulsive therapy and other physical treatments, such as insulin coma therapy and psychosurgery, the predominant view in biological psychiatry was that the major forms of mental illness were due to degenerative brain diseases, unyielding to therapeutic intervention. The early experience with convulsive therapy challenged this therapeutic nihilism and presaged the introduction of psychopharmacological agents.

Electrical induction of the generalized seizure soon became the preferred method, and electroconvulsive therapy (ECT) was the mainstay of biological treatment in psychiatry through the 1940s and 1950s. It was widely applied and central facts about the treatment emerged, such as its impressive efficacy in mood disorders (24). Nonetheless, most of the information regarding ECT was anecdotal, as modern standards for clinical trial methodology only emerged with the introduction of psychopharmacological agents. While initially ECT was the "gold standard" used to test the efficacy of the new psychotropics (29, 51), research in therapeutics soon focused almost exclusively on the medications. At the same time, particularly in the United States, the clinical use of ECT diminished (65), because it was apparent that pharmacological strategies often were efficacious and free of the cognitive side effects associated with ECT. In addition, with ECT perceived as the most invasive of the commonly used treatments in psychiatry, it became the bell weather for ideological debate about the biological bases of mental illness and the role of biological interventions. This in turn led to a virtual abandonment of ECT in some countries and a sharp reduction in its availability in public sector hospitals within the United States.

Spurred by growing awareness of the limitations of pharmacological approaches, as reflected in medication resistance, medication intolerance, safety concerns, and persistent side effects (e.g., tardive dyskinesia), a new era of research in ECT began in the late 1970s. This work incorporated more exacting methodological standards and led to new information about indications, treatment technique, side effects, and mechanisms of action. The role of ECT in therapeutics was the subject of a National Institutes of Health Consensus Conference (37) and its use was supported by a variety of national psychiatric and medical organizations (2, 48). In 1990, the American Psychiatric Association Task Force on ECT issued a report that represents a particularly comprehensive statement of standards of care (2).

Within the United States, there was a sharp decline in the use of ECT from the 1960s to 1980s (65). After 1980, rates of utilization stabilized and may have increased in recent years (66). ECT is administered to a far greater extent in private hospitals and academic centers than in public sector facilities. Indeed, a NIMH survey found that in 1980 not a single, non-white patient received ECT in a state facility in the United States (65). In contrast to the claims by opponents that the treatment is a method of behavior control inflicted on the destitute, the evidence indicates that, in general, the treatment is more readily available among the affluent.

Recent surveys indicate that within the United States approximately 80% of patients who receive ECT present with major depression (65, 66). Schizophrenia and mania are the next most common indications. Because of its use in major depression, females are more likely to receive ECT than males. The elderly also represent a high percentage of ECT recipients, presumably because ECT has a superior medical safety profile compared to some pharmacological alternatives and because rates of medication resistance and intolerance are elevated among the elderly (53).

The short-term efficacy of ECT in major depression is well established. Current recommendations suggest that ECT is an effective treatment for all subtypes of unipolar and bipolar major depression (2). In the absence of major depression, its use in the treatment of dysthymia has not been investigated and is generally discouraged. Electroconvulsive therapy is recommended as a treatment of first choice when medical or psychiatric considerations dictate a particular need for a rapid or definitive clinical response, when the risks of other treatments outweigh those of ECT, when there is a clear history of medication resistance and/or a history of favorable ECT response, or when patients indicate a preference for this modality. Increasingly, patients with major depression receive ECT after failing one or more adequate antidepressant medication trials during the index episode (45). There is considerable variability in practice about when the use of ECT is considered during the course of somatic treatment. Relevant issues here include the benefit of identifying a successful pharmacological approach that may then also be used as continuation therapy to prevent relapse, compared to the risk of unsuccessful pharmacological treatment prolonging the index episode (see Standard Antidepressant Pharmacotherapy for the Acute Treatment of Mood Disorders).

Within a few years of the introduction of ECT, it was recognized that therapeutic results in depressive illness were striking and often superior to those in schizophrenia (24). In the prepharmacological era, a large number of uncontrolled studies of depressed patients reported response rates of between 80% and 100% (2, 14). Post (43) suggested that prior to the introduction of ECT, the elderly depressed often manifested a chronic course or died of intercurrent medical illnesses in psychiatric institutions. A number of studies have contrasted the clinical outcome of depressed patients who received inadequate or no biological treatment to that of patients who received ECT. Although none of this work involved prospective, random assignment designs, the findings have been largely uniform. Electroconvulsive therapy resulted in decreased chronicity, decreased morbidity, and possibly decreased rates of mortality (53). In much of this work, the advantages of ECT were particularly pronounced in elderly populations.

Electroconvulsive therapy is a highly ritualized treatment, involving a complex, repeatedly administered procedure that is accompanied by high expectations of therapeutic success. Such conditions could augment placebo effects. Given this concern, a set of double-blind, random assignment trials were conducted in England during the late 1970s and 1980s that contrasted real ECT with that of the repeated administration of anesthesia alone—sham ECT. With one exception, real ECT was found consistently to be more efficacious than sham treatment (see ref. 51 for a review). The exceptional study (31) used a form of real ECT now known to have limited efficacy (54). Overall, these studies demonstrated that the passage of an electrical stimulus and/or the elicitation of a generalized seizure was necessary for ECT to exert antidepressant effects. Furthermore, the use of a repeated administration of anesthesia as a sham or active placebo condition may have underestimated therapeutic effects. There is some concern that repeated administration of anesthesia, by itself, may have mild antidepressant properties (51).

Following the introduction of tricyclic antidepressants (TCAs) and monoamine oxidase inhibitors (MAOIs), ECT was often used as the gold standard by which to calibrate the efficacy of the newer treatments (23, 51). Janicak et al. (53), in a meta-analysis of this work, reported that the average response rate to ECT was 20% higher when compared to TCAs and 45% higher when compared to MAOIs. No study has ever found a pharmacological agent to be superior in antidepressant effects when compared to ECT. Rather, ECT has been found to have either equal or superior efficacy. At the same time, this literature was characterized by major methodological limitations. Most critically, the era in which the bulk of these trials were conducted had different standards regarding the dose and duration of adequate pharmacological treatment. By modern standards, the pharmacological treatments used were often suboptimal (45, 57). In addition, given that patients must recognize whether they are receiving a medication or ECT, such comparisons could not be double blind.

Other issues intrinsic to the comparison of ECT with other biological treatments concern the speed and quality of clinical response. Since ECT is most often administered to inpatients, with severe and often psychotic depressive illness, rapid and definitive response is a fundamental concern. In the United States, the average course of ECT involves approximately eight to nine treatments, administered thrice weekly. This 3-week period to achieve full response indirectly supports the belief that symptomatic improvement is often more rapid with ECT compared to standard pharmacological treatments. However, only one recent study has tested this view (10). Patients who had failed an adequate TCA trial were randomized to treatment with ECT or lithium augmentation of the TCA. While both treatment conditions had equivalent efficacy, speed of improvement was more rapid with the TCA-lithium combination. Similarly, the extent of residual symptomatology following pharmacological treatment and ECT has rarely been compared. Hamilton (21), in a naturalistic, open study, reported that a higher percentage of depressed patients became asymptomatic after ECT compared to a similar group who received TCAs.

In recent years, the bulk of clinical trials involving ECT have focused on how variations in treatment technique impact on efficacy and side effects (58, 69). Historically, such work has provided critical information that has constrained theories of mechanisms of action and, recently such research has provided compelling controlled data regarding the overall efficacy of ECT.

The real–sham studies of the 1980s demonstrated that the passage of an electrical stimulus and/or the elicitation of a seizure were fundamental to antidepressant effects. Earlier research had underscored the importance of seizure elicitation. One set of studies found inferior clinical outcome when patients were administered subconvulsive electrical stimuli, as opposed to electrical stimuli of sufficient intensity to elicit generalized seizures (67). Another set of comparisons suggested that seizures elicited by chemical induction (e.g., flurothyl) were equivalent, if not superior, in therapeutic properties to those elicited with an electrical stimulus (50, 51). Furthermore, chemically induced seizures appeared to produce fewer cognitive side effects. This work further tied antidepressant effects to seizure induction, whereas adverse cognitive effects were related to the passage of an electrical stimulus. In line with this, Ottosson (41) performed a critical experiment, comparing three groups of patients. One group received standard suprathreshold electrical stimulation. Another group received the same electrical dosage, but with simultaneous administration of lidocaine, which, in addition to its antiarrhythmic properties, is an anticonvulsant and markedly reduced both the duration and amplitude of ictal activity. The third group received standard ECT, but with a substantially higher electrical intensity. The use of lidocaine to suppress expression of the seizure discharge resulted in diminished efficacy, whereas the high electrical intensity condition was associated with aggravated acute cognitive side effects. Consequently, this body of work lead to the view that, regardless of method, the elicitation of generalized seizures of adequate duration provides the necessary and sufficient conditions for antidepressant effects (14, 37). In contrast, the intensity of the electrical stimulus does not impact on efficacy, but contributes to the magnitude of cognitive side effects.

In recent years, these basic tenets have been substantially revised. The Columbia University group demonstrated that at low electrical intensity, generalized seizures of adequate duration are reliably produced with right unilateral ECT, but with remarkably weak therapeutic properties (54). In contrast, at low electrical dosage, bilateral ECT remained an efficacious treatment. Robin and de Tissera (47), using bilateral ECT, compared three different electrical waveforms that differed in overall stimulus intensity. They found that speed of clinical response was slower with less intense stimulation. Conjointly these findings led to the hypotheses that electrical dosage determines the efficacy of right unilateral ECT and influences speed of response, regardless of electrode placement. Both of these hypotheses were confirmed in a recent, random assignment, double-blind study by the Columbia University group (58). Crossing the factors of stimulus dosage (low or high) and electrode placement (right unilateral or bilateral), they found that short-term response rates to ECT varied between 17% and 65%. In addition, this work suggested that the effects of stimulus intensity on efficacy and efficiency (speed of response) were not related to the absolute electrical dose administered, but to the degree that the absolute dose exceeded the individual patient's seizure threshold (55, 58).

These new studies contradicted the fundamental tenets on mechanisms of action and offered a new perspective. Generalized seizures of adequate duration can be reliably produced that lack therapeutic properties. Therefore, the generalized seizure may be a necessary, but insufficient, condition for efficacy. Furthermore, there are electrical dose–response relationships in ECT, indicating that technical aspects in how the treatment is performed can have a major impact on efficacy. The interactions of electrode placement and electrical dosage in determining efficacy also suggest that antidepressant effects are dependent on current paths, implicating anatomic specificity in the functional systems therapeutically altered by ECT. With regard to broader considerations of efficacy, the new studies went beyond the real–sham ECT trials. Using double-blind methods in which each patient not only underwent repeated anesthetic administration but had generalized seizures of adequate duration, this work demonstrated that technical factors in treatment delivery (stimulus dosage and electrode placement) can have a dramatic effect on response rates (54, 58). Consequently, the neurobiological consequences of manipulating these parameters should be fundamental in accounting for antidepressant effects.

In the 1950s and 1960s, a series of studies showed impressive power to predict clinical outcome in depressed patients on the basis of pre-ECT symptomatological features and histories (see refs. 2 and 38 for reviews). This work is largely of historical interest, because the changes in nosology have resulted in more diagnosticly homogeneous samples being referred for ECT. Indeed, whereas the early research had emphasized the importance of vegetative or melancholic features as prognostic of positive outcome, recent studies restricted to patients with major depression suggest that subtyping as endogenous or melancholic has little predictive value (1, 2, 38). It appears that the early positive associations derived from the inclusion of neurotic depressives or dysthymics in the sampling.

In recent research, a few clinical features have been related to short-term ECT outcome. Several studies have noted that patients with a long duration of their current episode are less likely to respond (38). The majority of studies that have examined the distinction between psychotic and nonpsychotic depression found superior response rates among the psychotic subtype (38). This is of particular note given the established inferior response rate of this subtype to monotherapy with antidepressant or neuroleptic medications. In addition, in a relatively large number of studies patient age was positively associated with superior ECT outcome (53).

Although the predictive relationships between ECT outcome and episode duration, psychosis, and age are of theoretical interest, they are of insufficient strength to guide treatment recommendations. Recently, an alternative approach has been suggested: Electroconvulsive therapy was introduced in the prepharmacological era and was commonly used as a treatment of first choice. Increasingly, however, depressed patients now referred for ECT have frequently failed adequate trials of antidepressant medications. Despite this major shift in the nature of patient populations, there has been little research examining whether expectations about ECT response rates require recalibration. Prudic et al. (45) reported that patients who failed one or more adequate TCA trials prior to ECT had a 50% rate of response to bilateral ECT compared to an 86% rate in patients not found to be medication resistant during the index episode. Furthermore, medication resistance was particularly common in patients whose current episode was of long duration and particularly uncommon in psychotic depression, with the findings suggesting that the predictive power of these clinical features may be due to their relations with medication history. There is a need to replicate and extend this approach to prediction of ECT outcome. For example, it is unknown whether failure to respond to adequate treatment with a serotonin reuptake inhibitor (SSRI) has the same predictive power as TCA failure, an issue of both clinical and theoretical importance.

There have also been a variety of attempts to examine biological measures as predictors of short-term ECT outcome, including levels of neurotransmitter metabolites, neuroendocrine measures, and electroencephalographic (EEG) and other brain-imaging parameters. To date, none of these approaches has yielded consistent findings (1, 38). In the past, it was believed that as long as generalized seizures were of sufficient duration, maximal therapeutic effects would be obtained. It is now recognized, however, that seizure duration bears little relation to efficacy (55) and that generalized seizures of adequate duration can be consistently produced that lack therapeutic properties. Therefore, the field lacks a marker to indicate that therapeutically optimal treatment is being delivered. Because ECT results in acute changes in the levels of a variety of transmitters and peptides, the predictive power of these acute changes have been examined with regard to clinical outcome. Most of this work has documented null results. An exception, however, are reports by Scott and colleagues (46) that the acute release of oxytocin-associated neurophysin following the first ECT is associated with ultimate clinical outcome. Recently, this group suggested that this effect may pertain more to the adequacy of treatment methods than to patient characteristics. They found that the release of oxytocin acutely following ECT was sensitive to stimulus intensity and presumed that this sensitivity accounted for the relationship with clinical outcome. Other recent research avenues have concentrated on providing electrophysiological markers of treatment adequacy. There are consistent findings that forms of ECT that differ in efficacy also differ in the EEG parameters during and immediately following the seizure (39). Ongoing work is examining the sensitivity and specificity of such effects relative to clinical outcome.

As in major depression, ECT has been used to treat acute manic episodes for more than five decades. A recent review compiling this experience indicated that approximately 80% of 589 manic patients achieved remission following ECT (36). In contrast, surveys of utilization indicate that the use of ECT in the treatment of acute mania is often unappreciated. Particularly when patients manifest manic delirium, ECT may be life saving.

The bulk of the literature on the use of ECT in mania involves uncontrolled case series. In addition, pharmacological intervention, particularly with lithium, is typically the first-choice treatment and there are limited data on the efficacy of ECT in medication-resistant manic patients. Further complicating the use of ECT in mania was the belief expressed in early reports that manic patients required particularly intense forms of ECT (more than one treatment per day) or prolonged ECT courses to achieve remission (15, 24). A related view was that ECT did not exert a primary antimanic effect but that apparent clinical improvement was due to the masking effects of a treatment-induced organic mental syndrome.

Some of the limitations of the uncontrolled case series were addressed in a group of retrospective and prospective controlled comparisons of ECT and pharmacological treatments. In retrospective comparisons of ECT with lithium and chlorpromazine, the short-term efficacy of ECT was found in one study to be equivalent to both lithium and chlorpromazine and in another study to be superior to chlorpromazine (36). There have been only two prospective studies of ECT in acute mania, and both involved relatively small samples. In one study, patients were randomly assigned to treatment with ECT or lithium, with a neuroleptic used adjunctively in both treatment groups (61). At the end of this 8-week trial, the ECT and lithium groups had equivalent response rates, but the differences observed during the trial in discrete aspects of psychopathology all favored the ECT group. The second prospective study selected only patients who had failed to respond to robust treatment with lithium or neuroleptics (36). Patients were randomized to various forms of ECT or to pharmacological treatment with combined lithium and haloperidol. Among these medication-resistant patients, there was an impressive rate of response to ECT, but none of the small group of patients in the combined pharmacology condition met response criteria.

The findings of the retrospective and prospective studies were consonant with those of the uncontrolled reports in indicating that ECT is effective in acute mania and may be of particular value in manic patients who do not benefit from traditional pharmacotherapy. It is noteworthy, however, that there has yet to be a comparison of the efficacy of ECT and anticonvulsant medications. There is also limited information on the efficacy of ECT in rapid cycling patients. The recent prospective studies suggested that manic patients with poor premorbid functioning or those with symptoms of anger and irritability may be less likely to respond to ECT (36). This work and the recent retrospective studies also contradicted the view that particularly intensive forms of ECT are required in acute mania. In these studies, ECT was administered at twice- or thrice-a-week schedules and the average number of treatments ranged from 5.4 to 11. Indeed, recent evidence suggests that speed of improvement is often more rapid in acute mania than in major depression. The prospective studies have also rejected the view that improvement with ECT in acute mania is secondary to the induction of persistent confusion. Shortly following the ECT course, manic patients who responded generally manifested improved cognitive functioning without evidence of an organic mental syndrome (36).

There is considerable controversy about the role of ECT in the treatment of schizophrenia. Surveys of utilization indicate marked disparity in rates of use between and within countries. Likewise, recommendations of expert groups and professional organizations have been contradictory. For example, an National Institutes of Health consensus panel stated that evidence regarding the efficacy of ECT in schizophrenia was not compelling (37). The American Psychiatric Association Task Force on ECT recommended that ECT be considered particularly for schizophrenic patients who manifest prominent affective features or catatonia during exacerbations (2). In contrast, the Royal College of Psychiatrists expressed skepticism that any symptomatological features were predictive of response to ECT in schizophrenia (48).

Many of the earlier reports on this use of ECT consisted of uncontrolled case material (29). Diagnostic practice preceded the introduction of operationalized criteria for schizophrenia and patient samples and outcome criteria were often poorly specified. International differences in the rate of diagnosis of schizophrenia have been described previously. In particular, earlier American studies describing samples of schizophrenic patients may have contained substantial representation of mood disorder patients, who may be particularly likely to respond to ECT.

Overall, the earlier reports were enthusiastic about the use of ECT for schizophrenic patients with relatively recent onset of illness, with recovery or marked improvement noted in a large proportion of cases, typically on the order of 75% (1, 14, 29). Historical comparisons and comparisons to psychotherapy or milieu therapy suggested that the introduction of ECT resulted in both superior short-term clinical outcome and more sustained remissions (29). However, in this early era, the view was also frequently expressed that ECT was considerably less effective in schizophrenic patients with insidious onset and long duration of illness (24). In addition, it was suggested that schizophrenic patients often required intensive courses of ECT, involving more frequent and closely spaced treatments (14, 24).

A surprising number of prospective, controlled studies have addressed the efficacy of ECT in schizophrenia. As in major depression, ECT was initially used as the gold standard by which to establish the efficacy of pharmacological agents, in this case, neuroleptics. This work included comparisons of real and sham ECT, monotherapy with ECT or neuroleptic medications, and the combined use of ECT and neuroleptics.

Using real–sham designs, four studies were conducted in the 1950s and 1960s, and three recent studies in the 1980s (29). With the possible exception of Ulett et al. (67), the early studies failed to demonstrate a therapeutic advantage for real ECT compared to repeated administration of anesthesia alone. In contrast, at least in the short-term, the three recent studies each found clinically significant therapeutic advantages for real over sham ECT (29). The source of this discrepancy is unknown. The recent studies have had small sample sizes, but this should have limited the possibility of obtaining statistically significant findings. It is noteworthy that Taylor and Fleminger (64) explicitly focused on a middle-prognostic group, and excluded patients with chronic conditions. The high representation of patients with chronic schizophrenia in the early studies may have mitigated against a real ECT advantage (29). Another possibility is that in each of the recent studies, patients were maintained on neuroleptic medications during the clinical trial. There is evidence that the combination of ECT and neuroleptics is a more effective treatment than either form of monotherapy (29).

In the recent work the advantages of real relative to sham ECT only pertained to the period of time during and immediately following the acute treatment course. Within months of trial termination, symptomatic differences between the groups were not evident. The importance of these negative findings is questionable. In each case, the treatment received following the randomized trial was uncontrolled. Indeed, in some cases, patients assigned to the sham condition went on to receive ECT.

Ten prospective, controlled trials compared the efficacy of ECT with that of monotherapy with neuroleptics and/or other somatic treatments (e.g., insulin coma, reserpine). Since 1980, only one such study has been reported. The limitations of this literature in fundamental aspects of clinical trial methodology, particularly the reliability and validity of diagnosis, the nature of assignment to treatment groups, treatment adequacy, and the blindness and reliability of clinical evaluations, underscore the need for caution in interpretation (29). With these caveats, it appears that short-term ECT outcome was generally inferior or equivalent to antipsychotic medication (29, 33), although there were exceptions (12)). There is little indication from this literature about the clinical or treatment history features that might distinguish schizophrenics who preferentially respond to antipsychotic medications or ECT. In contrast, a seemingly consistent and surprising theme in this literature was the suggestion that patients who were administered ECT had superior long-term outcome compared to medication groups (12, 33). This pattern emerged in other retrospective studies (29), as well as in the most rigorous of the prospective controlled investigations, the work conducted by May and colleagues (33). This type of effect is unexpected, given the predominant perspective that virtually all the behavioral and physiological effects of ECT are relatively short-lived (50). Furthermore, most of these studies were conducted in an era when the importance of continuation and maintenance treatment was not appreciated and no study controlled the treatment received following resolution of the schizophrenic episode. Nonetheless, the possibility that ECT may have beneficial long-term effects merits attention.

There have been three prospective clinical trials comparing the efficacy of combination treatment with ECT and neuroleptics with that of ECT alone, and seven such trials compare the combination treatment to monotherapy with a neuroleptic (29). This literature is also characterized by a host of methodological problems. Relatively few of these studies involved random assignment, and fewer still involved fully blind assessment of treatment outcome. Nonetheless, it is noteworthy that in each of the three studies in which ECT alone was compared to ECT combined with antipsychotic medication there were indications that the combination was more effective (29). With one exception, in each of seven comparisons, the combination of ECT with antipsychotic medication was more effective than treatment with an antipsychotic medication alone. In some cases, a superior outcome was obtained despite lower average neuroleptic dose in the combination condition. This pattern is particularly impressive when one considers that monotherapy with neuroleptics has established efficacy and the resultant constraints on statistical power in establishing superior response to a combination condition. At the least, these studies suggest that neuroleptic medication should not be discontinued when schizophrenic patients are referred for ECT. Few of these studies followed patients beyond the acute treatment period, and in none was there standardization of continuation or maintenance treatments. Therefore, the relative persistence of any advantage for combination treatment is unknown. Nonetheless, the findings in some of these studies suggested a lower relapse rate in patients treated acutely with neuroleptics and ECT relative to neuroleptics alone. When considered with the follow-up results of May et al. (33), there is added reason to explore whether ECT, particularly in combination with antipsychotic medication, exerts a long-term beneficial effect in schizophrenia.

Antipsychotic medication is the first-choice treatment in schizophrenia, and ECT is typically reserved for medication-resistant patients. There have been eight, largely impressionistic reports on the use of ECT in such patients. We have yet to have a double-blind, random assignment study contrasting the efficacy of combined ECT and neuroleptic treatment with continued neuroleptic treatment alone in medication-resistant schizophrenic patients. Nonetheless, starting in the early 1960s there have been a series of reports that some medication-resistant schizophrenic patients benefit substantially by the addition of ECT (29). In some cases these positive reports contradict the clinical tenet that ECT is of limited value in chronic schizophrenic patients, with long durations of illness (1, 24), and there are a variety of unanswered questions. It could be that the recent set of positive case series contained samples with high representation of patients with prominent affective symptomatology (29). It could be that, in the absence of affective symptoms, the combination of ECT and neuroleptic treatment is particularly valuable for medication-resistant patients who have relatively short duration of illness. Regardless, it is clear that better information is needed, particularly about the clinical and historical features of those medication-resistant patients who may benefit from combination treatment. The patient that fails neuroleptic treatment and clozapine presents a serious treatment dilemma. It is noteworthy that the literature on the combined use of clozapine and ECT is confined to small case series (29).

Early studies of ECT suggested that the features associated with positive short-term clinical outcome in schizophrenia included being married, having at least a skilled or clerical occupation, an absence of premorbid personality disturbance and poor premorbid functioning, manifestation of catatonic symptoms, affective symptoms, and, most commonly, acute onset and short illness duration (29). It is noteworthy that many of these features have been found to predict outcome with pharmacological treatment and may be more general markers of prognosis. Nonetheless, the validity of these findings is questionable given that the early studies often used assessment techniques of doubtful reliability and lacked standardized diagnostic criteria. Although more recent research has replicated relationships between chronicity and outcome, predictive relationships were not observed for the presence of affective or catatonic symptoms (11, 29).

The potential for acute extrapyramidal syndromes, particularly neuroleptic-induced parkinsonism (NIP), and for persistent tardive dyskinesia are major drawbacks of traditional neuroleptic treatment. Although extrapyramidal symptoms are generally considered to be reversible, prospective studies have suggested that neuroleptic-induced parkinsonism may predict the subsequent development of tardive dyskinesia.

Because ECT is unique in having both antipsychotic and antiparkinsonian properties, it may exert ameliorative effects on neuroleptic-induced parkinsonism (1, 29, 19). For example, Goswami et al. (19) studied nine schizophrenic inpatients with a longitudinal triphasic design, first using neuroleptics, then neuroleptics and ECT, and then neuroleptics. Neuroleptic-induced parkinsonism was significantly reduced in stepwise fashion when patients were treated with ECT. Recently, Mukherjee and Debsikdar (35) introduced the notion that ECT may protect against the later development of neuroleptic-induced parkinsonism and tardive dyskinesia. They examined 35 DSM-IIIR schizophrenic patients who were on neuroleptics for at least 2 weeks, all of whom were receiving (n = 15) or had received (n = 20) a course of unmodified bilateral ECT during the index episode. None of the patients had bradykinesia, rigidity, or postural instability and only one patient met the research diagnostic criteria for probable tardive dyskinesia. Mukherjee and Debsikdar (35) speculated that if neuroleptic-induced parkinsonism is a risk factor in the development of tardive dyskinesia, ECT may ultimately protect against tardive dyskinesia by preventing initial neuroleptic-induced parkinsonism. At the neurophysiological level, there is evidence in rodents that electroconvulsive shock (ECS) prevents the development of dopamine receptor supersensitivity with exposure to dopamine antagonists (29) and results in increased D1 receptor density without impact on the D2 receptor density (16).

There are also suggestions that a history of ECT may be associated with a low prevalence or delayed development of tardive dyskinesia. Gardos et al. (18) evaluated 122 schizophrenic outpatients in Hungary and reported a striking absence of severe tardive dyskinesia. They suggested that the low prevalence was due to the use of ECT to treat exacerbations and the avoidance of high dosage neuroleptic treatment. In an American sample, Cole et al. (3) recently reported that a history of ECT was associated with a lower risk and delayed appearance of tardive dyskinesia. As noted, Mukherjee and Debsikdar (35) found virtually no tardive dyskinesia in an Indian sample, which they attributed to the use of ECT. Schwartz et al. (59), in an Israeli sample, reported a reduced incidence of tardive dyskinesia among male schizophrenic patients with a history of ECT. If, in fact, ECT does offer long-term protection against the iatrogenic effects of concurrent or later exposure to neuroleptics, this would also contradict the general impression that the behavioral and physiological effects of ECT are uniformly transient (50).

There is limited information on the use of ECT in other psychiatric disorders. Among the anxiety disorders, traditionally it has been felt that ECT lacks efficacy in obsessive–compulsive disorder (OCD) and generalized anxiety disorder (14). Aside from clinical implications, the putative lack of efficacy of ECT in OCD is of theoretical interest, given the established efficacy of antidepressant medication, particularly those with strong serotonergic effects. Only one recent study examined this issue. Khanna et al. (26) reported on a small, prospective series of nondepressed OCD patients who had failed trials of antidepressant medications and behavior therapy. Shortly following ECT, marked improvement was documented, but patients were not given continuation treatment and quickly relapsed. Antidepressant medications are also efficacious in panic disorder. Surprisingly, there is virtually no information on the effects of ECT on panic disorder, even in patients with comorbid major depression.

Catatonia may be manifested in a variety of psychiatric disorders or as a consequence of medical illness. Impressionistic data support the use of ECT in catatonia, particularly following poor response to treatment with benzodiazepines (42). The syndrome of lethal catatonia is a life-threatening condition, characterized by stupor or excitement, hyperthermia, clouded consciousness, and autonomic dysregulation (32). Mann et al. (32) identified 292 cases in the world literature, with deaths in 176 (60%) of these cases. The literature on this syndrome, which consists solely of case series, suggests that neuroleptic treatment is of limited efficacy. Given the difficulty in distinguishing lethal catatonia from neuroleptic malignant syndrome (NMS), escalation of neuroleptic dosage may be counterproductive. Particularly when instituted prior to a comatose stage, ECT appears to be effective (29). In a recent study of schizophrenic and schizoaffective patients, perplexity was a clinical sign associated with positive ECT outcome (11). Confusion or perplexity is a common feature of the European categorization of cycloid psychoses, which may be a variant of schizoaffective disorder, that traditionally has been thought to be exquisitely responsive to ECT.

A variety of organic affective and psychotic states, as well as certain states of delirium, have shown rapid and favorable response to ECT. These include delirium associated with alcohol withdrawal, toxic delirium or psychosis associated with use of phencyclidine, and mental syndromes associated with lupus erythematosus and enteric fevers (2). However, because of the availability of alternative treatments, ECT is rarely used in such cases.

In both open and sham-controlled trials, ECT has been found to improve clinical symptoms in idiopathic Parkinson's disease, at least on a short-term basis (1, 2, 29). Typically, L-dopa requirements are sharply reduced when parkinsonian patients receive ECT. The persistence of the beneficial effects are unpredictable and highly variable. The utility of ECT as a long-term treatment for medication-resistant Parkinson's disease has yet to be tested, as use of maintenance ECT has not been evaluated.

NMS shares clinical features with lethal catatonia and has been considered to be an iatrogenic form of lethal catatonia induced by neuroleptics (32). When the clinical community became cognizant of NMS, there was reluctance to treat these patients with ECT, based on the fact that NMS has similar symptoms to malignant hyperthermia, a familial syndrome provoked by exposure to general anesthesia and depolarizing muscle relaxants, such as succinylcholine. However, NMS and malignant hyperthermia have been shown to be unrelated syndromes and several reviews have documented that ECT is an effective treatment for NMS (e.g., see ref. 6). Davis et al. (6) found that mortality rates in NMS patients were equivalent with ECT compared to bromocriptine, dantrolene, L-dopa, or amantadine, and averaged approximately half that of untreated patients. The complications and deaths that have been observed with use of ECT have been tied to continued administration of neuroleptic medication and to cardiac dysregulation. Given the marked hemodynamic alterations associated with ECT in the NMS patient, it is advisable to first use medication strategies to stabilize autonomic function before starting ECT. These patients must often be discontinued from antipsychotic medication; therefore, ECT may have the advantage of treating both the NMS and the underlying psychotic condition.

In a variety of animal models and in humans, ECT exerts pronounced anticonvulsant effects (44, 55). There is a case literature suggesting efficacy in some patients with medication-resistant epilepsy and instances of prolonged status epilepticus (2). The possible utility of ECT in status epilepticus may be of clinical relevance, because the mortality rate with standard pharmacological protocols remains high.

Electroconvulsive therapy is most commonly used as a short-term treatment for acute episodes or exacerbations of psychiatric illness. In major depression, it is established that relapse rates during the first 6 months following response to ECT will be high if continuation treatment is not provided (57). A similar pattern is expected in mania and schizophrenia. For depressed patients, it has long been thought that relapse rates can be markedly reduced by the use of classical antidepressant medications, specifically TCAs or MAOIs, as continuation treatment following response to ECT. This view was based on the results of three double-blind, placebo controlled, randomized trials conducting in England during the 1960s (57). This work focused on patients receiving ECT as a first-choice treatment. Since then, ECT samples increasingly contain patients who fail to respond to these medications during the acute episode. Theoretically, if an agent that proved ineffective in treating the acute episode was effective as a continuation or maintenance therapy following ECT, then either of two hypotheses would be supported. From a neurobiological perspective, this might suggest that what needs to be accomplished to prevent relapse differs in degree or kind from what is required to treat the acute episode. Alternatively, ECT may change the neurobiology so that a medication can exert an action as a continuation treatment that it could not exert during pre-ECT acute-phase treatment. However, recent studies document high relapse rates following response to ECT in major depression in patients receiving traditional continuation pharmacotherapy (57). Furthermore, there is preliminary evidence that patients with established resistance to adequate trials of antidepressant medications during the acute episode have particularly elevated relapse rates and that adequate continuation therapy with a TCA may only benefit those who did not fail this class of medication during the acute episode (57). Consequently, there is considerable uncertainty about optimal continuation pharmacotherapy following ECT and a clear need for controlled research to reevaluate traditional and alternative strategies. Such efforts are in progress.

The problem of relapse is likely related to the fact that ECT is the only somatic treatment in psychiatry that is typically discontinued once it has been shown to be effective. An alternative is to use ECT as a form of continuation or maintenance treatment (2). There is a long history of such a use of ECT in both mood disorders and schizophrenia. On the one hand, this impressionistic literature suggests that, continuation ECT may be broadly effective; however there are also problems with its acceptability to patients and with compliance. Furthermore, no prospective, controlled study has compared the efficacy of continuation ECT with that of continuation pharmacotherapy. Research of this type is also indicated.

The mortality rate associated with ECT is comparable to that of general anesthesia in minor surgery and estimated to be approximately one death per 10,000 patients treated (1, 2). Patients believed to have increased risk of morbidity are those with space-occupying cerebral lesions or other conditions that increase intracranial pressure, recent myocardial infarction associated with unstable cardiac function, recent intracerebral hemorrhage, unstable vascular aneurysm or malformation, retinal detachment, pheochromocytoma, or any patient rated at ASA level 4 or 5. The prevalence of these conditions increases with aging (53). There is strong consistency in the literature documenting that ECT-related medical complications are more likely in the elderly, particularly the oldest age groups, in those with preexisting medical conditions, particularly cardiac illness, and in those receiving concurrent medication for medical conditions (53). Electroconvulsive therapy is often described as safer than classical antidepressant medications, particularly among the frail elderly. There are few controlled comparisons that have tested this claim (70), and naturalistic data are compromised by the fact that high-risk patients are preferentially referred for ECT.

The application of an electrical stimulus results in immediate vagal stimulation, parasympathetic outflow, and bradycardia. With the elicitation of a generalized seizure, there is outpouring of catecholamines, markedly enhanced sympathetic tone, and resultant tachycardia and hypertension. The magnitude of the hemodynamic changes is often great, but short-lived, with a return to baseline levels within minutes of seizure termination. Benign arrhythmias are common in the immediate postictal period (2, 70).

Mortality or significant morbidity during a course of ECT typically occurs immediately following the seizure or during the postictal recovery period. Cardiovascular complications are the leading cause of such events (2, 34). Controlled investigation has shown that the likelihood of cardiovascular complications is substantially increased in patients with preexisting cardiac illness and that the type of preexisting illness strongly predicts the nature of the complication, that is, ischemic disease or ventricular arrhythmia (70). In recent years, there has been considerable variability among practitioners in the prophylactic use of b-adrenergic blocking agents, such as labetalol or esmolol, to reduce the hypertensive effects of seizure induction (34). Whether this strategy is effective in limiting cardiovascular morbidity has not been documented, and there is a theoretical concern that this approach may be counterproductive. The increased cardiac output and peripheral hypertension associated with the seizure may contribute to the profound ictal increase in cerebral blood flow (CBF). In turn, enhanced CBF provides for the transport of the oxygen and carbohydrate supplies that are necessary to sustain the large ictal increase in cerebral metabolic rate (22). Indeed, it is established that some of the b-adrenergic blocking agents, and newer anesthetics, like propofol, that reduce ictal hypertension, are also associated with shortened seizure duration (34, 55). Prospective investigation is needed of the efficacy and safety of pharmacological strategies that limit the cardiovascular morbidity of ECT.

Despite increased CBF and intracranial pressure, cerebrovascular events are extremely rare among patients receiving ECT. Other sources of morbidity are prolonged and tardive seizures. The risk of prolonged seizures or status epilepticus is increased by some concomitant pharmacological treatments. These include use of adenosine antagonists, such as theophylline, high dosage of some neuroleptics, and lithium (2, 55). This risk may also be increased among patients with a preexisting electrolyte imbalance and when more than one seizure is induced in the same treatment session (2).

When the phenomenon of kindled seizures in animals was first demonstrated, concern was expressed that ECT may result in kindling, producing a persistent decrease of seizure threshold and creating a vulnerability for the later development of seizure disorder (28). However, ECS in animals exerts an anticonvulsant effect, blocking the development or expression of kindling (44, 55). For the most part, epidemiological studies of the frequency of spontaneous seizures and related EEG phenomena in former ECT patients have also not supported this concern (28). Recently, no relation with history of ECT was found in quantitative studies of human seizure threshold (28).

The cognitive effects of ECT are the major factor limiting its use. These effects are stereotyped, are expected to be independent of psychiatric diagnosis, and are well modeled in animal studies (52). Central facts to be kept in mind when considering the cognitive consequences of ECT concern the slope of recovery functions, the impact of technical factors in treatment administration, and individual differences among patients.

In most cognitive domains, there is rapid recovery of function shortly following ECT (52). Consequently, the type and magnitude of cognitive deficits are heavily determined by when they are observed (4, 58). When additive, progressive deficits occur during the ECT course, this is partly attributable to the spacing of treatments, with incomplete recovery before the imposition of further treatment (4).

How ECT is performed strongly impacts on the breadth and severity of short-term cognitive deficits. These factors include the anatomic positioning of electrodes, the type of electrical waveform used, the intensity of the electrical stimulus, and the spacing of treatments (52, 58, 69). Compared to application of electrodes over the right hemisphere (right unilateral ECT), the traditional bifrontotemporal placement of electrodes (bilateral ECT) results in prolonged recovery of orientation during the postictal period, a greater probability of developing an organic mental syndrome, and more extensive and severe amnesic effects. The fact that left and right unilateral ECT differ in the extent of amnesia they produce for verbal rather than nonverbal material clearly indicates lateralized neurophysiological effects (52). Electrical waveforms that are inefficient in seizure-eliciting properties (e.g., sine wave) produce greater cognitive disturbance than more efficient waveforms (brief and ultrabrief pulse) (52, 69). Similarly, there is evidence that, within a waveform, the extent to which electrical dosage exceeds seizure threshold contributes to the magnitude of short-term cognitive side effects (55, 58).

Patients also vary considerably in the extent and severity of cognitive side effects. There are limited data on the individual factors that predict this vulnerability. A set of mainly retrospective studies has suggested that advanced age heightens the probability of prolonged confusion and that this effect is most pronounced among the very old (53). In addition, receiving psychotropic medication during ECT or experiencing major medical illnesses prior to ECT appear to be additional risk factors, particularly among the elderly. However, in much of this work, the manner in which ECT was conducted was not optimal. This problem is fundamental, because, depending on ECT technique, at some centers more than 50% of elderly patients have been found to develop an organic mental syndrome and elsewhere virtually no patient develops this adverse effect (58). It is commonly suggested that preexisting cognitive deficit is also a risk factor for more pronounced ECT-induced cognitive effects (2). Here as well, the data are extremely limited. Retrospective studies of patients with poststroke depression, depression and dementia, and depression and Parkinsonism, and other neurological disorders have generally reported favorable clinical outcome, with only a suggestion of increased tendency for prolonged postictal confusion (53). This issue is complicated by the fact that depressed patients with pseudodementia typically show progressive improvement in cognitive function during the ECT course. Recently, it has been suggested that manifestation of T2-weighted MRI signal abnormalities in the basal ganglia is associated with post-ECT delirium (13).

In the immediate postictal period, patients may manifest, as they do with spontaneous seizures, transient neurological abnormalities, alterations of consciousness [e.g., disorientation or attentional dysfunction], sensorimotor abnormalities, and disturbance in higher cognitive functions, particularly learning and memory (52). The severity and persistence of these acute effects show marked sensitivity to technical factors in ECT administration. These factors determine, for instance, whether patients require a few minutes to achieve full reorientation following seizure elicitation or several hours (4, 52, 58).

Recovery of cognitive function following a single treatment is rapid. However, with forms of ECT that exert more severe acute cognitive effects, recovery may be incomplete by the time of the next treatment. In such cases, repeated acute assessment at the same time points relative to seizure induction demonstrates deterioration over the treatment course (4, 52). In some cases, patients may develop an organic mental syndrome with marked disorientation during the ECT course (13, 52). Milder forms of ECT have been developed in which cumulative deterioration in cognitive functions does not occur. Indeed, with specific alterations of ECT technique, cumulative improvement in some acute cognitive measures has been demonstrated (52).

Associations between the magnitude of cognitive effects and ECT treatment parameters rapidly diminish as time from ECT progresses. More than a week or two following the end of the ECT course, differences between electrode placements are difficult to establish (52, 58, 69). Within days of the end of the ECT course, depressed patients manifest performance superior to their pretreatment baseline in most cognitive domains. For example, patients typically show superior IQ scores on tests of intelligence shortly following ECT relative to those in the untreated depressed state (52). Similarly, prior to treatment, depressed patients often manifest deficits in the acquisition of information, as revealed by tests of immediate recall or recognition of item lists. Within several days following the ECT course, patients are typically unchanged or improved in the immediate memory measures, with the change in clinical state being the critical predictor of the magnitude of improvement. In contrast, patients often manifest a marked disturbance in their ability to retain information over a delay. This reflects a double dissociation between the effects of depression and ECT on anterograde learning and memory (52). Depression is associated with an acquisition deficit, most likely related to disturbances in attention and concentration, which frequently recedes following treatment with ECT. In contrast, ECT introduces a new deficit in consolidation or retention, so that information that is newly learned is rapidly forgotten.

During and shortly following a course of ECT, patients also display retrograde amnesia. Deficits in the recall or recognition of both personal and public information are usually evident, and there is evidence that these deficits are greatest for events that occurred temporally closest to the treatment (63). Therefore, memory for more remote events is usually intact, but patients may have difficulty recalling events that transpired during and several months to years prior to the ECT course. The retrograde amnesia is rarely dense, as patients typically show spottiness in memory for recent events. As time from treatment increases, there is typically improved retrograde functioning, with a return of more distant memories. This temporally graded pattern indicating greatest vulnerability for more recent events is compatible with similar findings of the effects of repeated ECS in animals (52). Both the anterograde and retrograde amnesia are most marked for explicit or declarative memory, whereas there should be no effect on implicit or procedural memory (52, 63).

Within a few weeks following the end of ECT, objective evidence of persistent cognitive deficit is difficult to document (52). The anterograde amnesia typically resolves within a few weeks of ECT termination (58). The retrograde amnesia often shows a more gradual reduction, with substantial return of memory for events that were seemingly forgotten immediately following the treatment course. However, persistent effects of ECT have been identified (69). Most likely because of a combination of retrograde and anterograde effects, patients may manifest, even when tested at substantial time periods after treatment, persistent amnesia for some events that transpired in the interval of several months prior to and following the ECT course (52, 69).

There has been concerted investigation of the perceptions of former patients regarding the effects of ECT on their cognitive functioning. Using a variety of instruments to elicit self-reports and to assess memory, investigators have not observed significant relations between changes in subjective and objective measures of memory functioning at any testing interval (52, 58, 69). With noted consistency, symptomatic severity has been found to strongly correlate with patients' evaluations of their memory functioning, both before and after ECT (52, 58, 69). Early studies by Squire (63), suggested that there was no change in the level of memory complaints shortly following right unilateral ECT, but that there was an increase in memory complaints and a redistribution in the nature of perceived deficits with bilateral ECT. When patients were tested several months after bilateral ECT, they showed a reduction of memory complaints, reaching the global level that was observed in patients after unilateral ECT; nonetheless, they showed a persistent redistribution in the nature of complaints. However, since this work, a series of studies has found that the great majority of depressed patients report fewer memory difficulties within a week of the ECT course than they had prior to the course. This improvement in subjective reports has been observed both for groups treated with right unilateral and bilateral ECT (52, 58, 69).

A small minority of former patients report profound and long-lasting cognitive impairment, which they attribute to this treatment modality. The validity of this phenomenon has not been established with objective testing (17) and the reasons for this discrepancy are not clear. Complicating this issue are the effects of current psychopathology on cognitive functions (17), the natural progression of some forms of psychiatric illness, the effects of current and past treatments other than ECT, and the methodological difficulties intrinsic in isolating a rare phenomenon. In addition, because of the public and scientific attention that has been given to the cognitive side effects of ECT, patients who receive this modality may be primed to attribute perceived changes in cognitive functioning to this treatment relative to other causes (63).

It is astonishing that repeated seizures in humans can exert such profound therapeutic effects. From a practical viewpoint, investigations of the mechanisms of action of ECT have the advantage of typically studying the most severely depressed patients who subsequently show marked improvement in clinical state. Furthermore, unlike pharmacological treatments, ECT is administered in a punctate manner, and the course of biological changes can be examined relative to the timing of the intervention. Because ECT is essentially a nonpharmacological physical treatment, patients can be studied while medicationfree both prior to and following the intervention, removing a major complication in many human studies of the mode of action of psychotropics. Finally, it is evident that the anatomic positioning of electrodes (bilateral or unilateral ECT) has marked impact on neurophysiological and cognitive sequelae, and, in some circumstances, on efficacy. This potential to restrict the anatomy of the sites involved in seizure initiation and propagation is presently limited by the use of an electrical stimulus and the intrinsic geometry and impedance properties of the scalp and skull (56). The development of new focal methods of brain stimulation and seizure induction will likely involve the use of time-varying, focused magnetic fields and may afford new opportunities to isolate the functional neural systems involved in antidepressant and other therapeutic effects (56).

Over a hundred theories have been offered to account for the efficacy of ECT, and skepticism has often been expressed about the possibility of uncovering its mode of action, in part because ECT produces a wide variety of effects on neurophysiological, neurotransmitter, and neuroendocrine systems. The provocation of generalized seizures results in so many systemic changes that isolating those critical to efficacy from epiphenomena was believed to be relatively hopeless (50). The demonstrations that generalized seizures can be reliably produced that lack therapeutic properties has engendered greater optimism. Using subtraction methods, preclinical and clinical research has begun to isolate neurobiological changes that accompany therapeutically effective forms of stimulation from nonspecific changes intrinsic to seizure induction (e.g., see ref. 39).

Hundreds of preclinical studies have compared the neurochemical effects of ECS and antidepressant medications. Some investigators have argued that a unitary mode of action underlies the efficacy of ECT and antidepressant medications, highlighting the common changes in noradrenergic and serotonergic transmission. Others have reached the opposite conclusion, emphasizing the disparities (15). Failing a firmer understanding of the pathophysiology of depressive illness or the bases of therapeutic response, information from preclinical investigation alone is unlikely to resolve this controversy.

At the clinical level, it is evident that ECT and antidepressant medications have different spectra of therapeutic action. The breadth of conditions that are posited as responsive to ECT from a single mode of action will strongly influence the search for mechanisms. For example, Post and colleagues (44) emphasized the efficacy of ECT in both depression and mania, and suggested parallels to lithium, carbamazepine, and valproate, supporting anticonvulsant theories of the mode of action of ECT (55). However, although it is clear that antidepressant medications (TCAs, MAOIs, and SSRIs) may aggravate mania, there is no assurance that the mode of action of ECT is uniform across major depression and mania.

An alternative strategy is to investigate the relationship between the response to a class of medications and to ECT (45, 57). It had long been assumed that among depressed patients, clinical outcome with ECT was independent of history of response to antidepressant medications (15). Were this the case, it would suggest at least partial independence in modes of action. For example, high rates of ECT response in patients clearly resistant to TCAs would suggest that ECT differs from TCAs in the degree or kind of therapeutic neurobiological changes induced. However, although it is evident that many TCA-resistant patients respond to ECT, the preliminary evidence suggests that their response rate is reduced compared to that of patients without established medication resistance (45). This, in turn, is compatible with some overlap in mechanisms of action.

One group of theories focuses on psychological factors as largely accounting for the antidepressant effects of ECT. Various theories have emphasized psychodynamic themes, viewing ECT is as a punitive intervention, the high expectations of recovery with ECT, focusing on the induction of placebo effects, or the adverse cognitive consequences of ECT, implicating amnesia or gross cognitive as the basis for perceived therapeutic response (50). The findings of the double-blind, random-assignment studies using real vs. sham designs rule out the psychodynamic and placebo explanations (51). Furthermore, the findings that, independent of seizure elicitation, the efficacy of ECT is contingent on technical factors in its administration, such as the anatomic positioning of electrodes and the intensity of electrical dosage, substantially contradict this kind of explanation (58). A large number of studies have sought relationships between cognitive and therapeutic changes. When significant relationships have been observed, invariably they have indicated better cognitive functioning in those patients with greatest improvement in symptomatic status (52).

The generalized seizures produced during ECT reflect the hypersynchronous discharge of large neuronal populations. They are hypermetabolic states, involving pronounced increases in CBF, oxygen consumption, and glucose metabolism. Typically, the magnitude of the CBF increase outstrips metabolic demands, and, with modified ECT changes are not expected in cerebral arterial-venous pO2, pCO2, or lactate gradients (22).

There is a refractory period immediately following electrically induced seizure in which seizure threshold is markedly elevated and repeated seizure induction is difficult (55). Spontaneous cortical electrical activity may also be suppressed immediately following seizure termination. Manifestation of this isoelectric electroencephalogram (EEG) pattern is more common with bilateral than with unilateral electrode placement and with high- than with low-intensity electrical stimulation; it also may be predictive of superior clinical outcome (39). During the postictal period, the isoelectric pattern is typically replaced by high-amplitude slow-wave (delta) activity. Likewise, pronounced cortical reductions in CBF and glucose metabolism have been demonstrated (Nobler et al., unpublished data). The increase in EEG slow-wave activity and the decrease in CBF and glucose metabolism are greatest in anterior cortical regions and when unilateral ECT is administered, in the hemisphere ipsilateral to electrode placement (68, Nobler et al., unpublished data).

These neurophysiological effects have been interpreted as reflecting the recruitment of endogenous inhibitory processes to terminate the seizure (55). Several of these anticonvulsant effects are progressive during the ECT course. Seizure threshold typically shows a cumulative increase, and there is a decrease in seizure duration (55). These two effects are at least partially independent, and pharmacological agents have been shown to have distinct impact on seizure threshold and duration (28, 55). Slow-wave EEG activity also shows progressive enhancement during the ECT course. There are preliminary data suggesting that the change in seizure threshold is related to efficacy in both major depression and mania (55). In particular, forms of ECT administration that exert weak antidepressant effects produce smaller increases in seizure threshold (58). However, a relationship between clinical outcome and changes in seizure duration has not been observed. A large number of studies have also sought to establish a relationship between changes in EEG slow-wave activity and clinical outcome (1, 68). These findings were inconclusive and largely limited by the fact that few investigations examined topographic effects. Furthermore, this work was largely conducted with high intensity forms of bilateral ECT, producing ceiling effects by inducing large increases in slow-wave activity across patients, possibly masking a relationship with treatment outcome. New findings suggest that greater CBF reductions in bilateral anterior cortical regions, as observed following both a single treatment and a complete course, are associated with superior antidepressant response (Nobler et al., unpublished data). To date, brain imaging studies of ECT have focused on cortical effects. The positive findings support the use of high-resolution technologies to outline the anatomy of the cortical–subcortical functional systems that subserve therapeutic response.

In man and other animals, electrical induction of generalized seizures results in a short-lasting reversible inhibition of protein synthesis (60). No comprehensive theory has related this effect to efficacy. However, there is a substantial body of evidence that suggests that medications that inhibit protein synthesis interfere with long-term memory (5). This may be relevant to the cognitive effects of ECT, suggesting disruption of the neuronal plasticity mechanisms necessary for memory consolidation. Alternatively, recent evidence suggests that unmodified ECS in rodents activates astrocytes, as measured by increased staining for glial fibrillary acid protein, and also leads to an increase in the ratio of a marker of newly formed synapses relative to a marker of mature synapses (27). This has been interpreted as suggesting that ECS results in increased synaptic remodeling, an effect that may be relevant to a variety of biochemical theories of mode of action.

There is evidence that in man and animals electrically induced seizures also result in transient disruption of the blood–brain barrier. This disruption is dependent on transient systemic hypertension and cerebral vasodilatation, with the presumed mechanism being increased vesicular transport by pinocytosis (22, 60). This effect is short lasting, and no increase in cerebrovascular permeability to serum proteins can be demonstrated by 24 hours after the last ECS in a series. Findings with repeat MRI in ECT patients of transient increases in T1 relaxation times, an index of water content in the brain, have been interpreted as reflecting this transient disruption.

In turn, ECT results in an acute, and typically transient, increase in the plasma of a variety of transmitters and peptides including epinephrine, norepinephrine, prolactin, b-endorphin immunoreactivity, vasopressin, oxytocin, adrenocorticotropin (ACTH), cortisol, insulin, follicle-stimulating hormone, and luteinizing hormone (50). Increased brain permeability to endogenous, circulating molecules with putative antidepressant effects has been raised frequently as a process related to efficacy (15). Furthermore, in schizophrenia, the seeming superiority of ECT combined with a neuroleptic to either treatment alone has been attributed to the enhanced permeability, which produces higher brain concentration of the neuroleptic. In contrast, with the exception of oxytocin-related neurophysin (46), findings relating the antidepressant effects of ECT to the magnitude of acute surges of endogenous substances in plasma have been negative.

The magnitude of the systemic hypertensive changes, a factor contributing to increased blood–brain barrier permeability, is also unrelated to clinical outcome. During the postictal period in schizophrenic patients, plasma and red blood cell levels of haloperidol have been shown to increase transiently, indicating a redistribution phenomenon. However, recent work has demonstrated a similar effect in rats but did not find any changes in cerebral concentrations of haloperidol (29).

Electroconvulsive shock has many consistent effects on transmitter and peptide concentrations, receptor density, and signal transduction mechanisms (see ref. 40 for a review). This evidence comes mainly from studies of physiologically normal rodents, and generalizability to human clinical context is uncertain. Criteria have been offered to screen ECS biochemical alterations for relevance to the mechanisms of action of ECT (40). For example, a single ECT treatment rarely results in clinical remission, and candidate neurochemical changes induced by ECS are expected to evolve with repeated application. However, critical aspects of methodology remain uncertain. Early work in which ECS was massed, with repeated application in the same day, was criticized as failing to mimic the schedule used in the human context, usually twice or thrice weekly. In turn, it is well known that there are fundamental differences between humans and rodents in the onset, peak, and recovery functions for a variety of biological processes. For example, following a single ECT treatment, seizure threshold in the human remains elevated at least for a matter of days. The elevation in the rat appears to last for about 1 hr (20). Studies of the neurochemistry subserving this anticonvulsant effect would have to take this difference in time course into account. Using such behavioral end points to guide parameter selection is largely unavailable in ECS studies aimed at accounting for the therapeutic effects of electrical induction of seizures.

In the human, in addition to neuroendocrine strategies, biochemical studies of ECT have concentrated on alterations in monoamine concentrations and turnover and receptor physiology in peripheral body fluids and tissue (49; see ref. 25 for a review). Here the relations to brain neurochemistry are often uncertain, in part due to concerns about the specificity of neuroendocrine probes or the strength of association between peripheral measures and central transmitter or peptide function. The development of in vivo brain imaging techniques to selectively probe central neurochemical systems may substantially advance knowledge in this area.

There are a number of effects that ECS has in common with chronic administration of TCAs in altering noradrenergic function. Repeated ECS results in a decrease in the density of b-adrenergic receptors (25, 40). This change appears to be an adaptive down-regulation, as microdialysis and other studies have shown prominent acute and basal increases in brain norepinephrine and tyrosine hydroxylase activity. The b-receptor reduction also appears to be functional, as it is associated with reduced receptor-mediated cyclic adenosine monophosphate production and behavioral response to b agonists are attenuated after ECS. The changes in norepinephrine (NE) concentration and b-receptor density appear to show regional specificity, being most marked in cortex and hippocampus, but not in striatum, cerebellum, or hypothalamus. In general, the evidence is less consistent with regard to a1- and a2-adrenergic receptors. There are discrepant reports regarding decreased a2-adrenergic receptor density, but behavioral data suggest presynaptic subsensitivity, perhaps accounting for increased basal NE concentrations. In the human, there is a surge in the levels of plasma catecholamines, particularly epinephrine, with seizure induction. These changes are transient and may be more relevant to ECT effects on cardiac function than to efficacy. With the possible exception of cerebrospinal fluid (CSF) homovanillic acid (HVA), studies of more chronic changes in peripheral or CSF measures of catecholamines and their metabolites have been negative or inconsistent (49). In general, the evidence is consistent with the view that repeated ECS enhances noradrenergic function.

The comparison of TCAs or MAOIs and ECS in altering serotonergic function is intriguing, since they have opposite effects on 5-HT2 receptor density (25, 40). Chronic ECS results in an increase, whereas antidepressant medications produce a decrease. Acutely following a single ECS, there is also increased serotonin (5-HT) concentrations, but no consistent effect has been observed on basal levels. However, as with some antidepressant medications, chronic ECS enhances electrophysiological and behavioral serotonergic responses mediated by the 5-HT2 receptor subtype. The effects of chronic ECS on 5-HT2 density and responsivity require intact serotonergic and noradrenergic innervation and can be blocked by inhibiting serotonin or NE synthesis. Studies of 5-HT1 receptor density and function have not reported theoretically consistent effects, but much of this work predated identification of specific 5-HT1 receptor subtypes. In humans, most studies have reported small or unchanged levels of plasma, urinary, and CSF 5-hydroxyindoleacetic acid (5-HIAA) following ECT, but a recent study with medication-free depressed patients suggests that CSF 5-HIAA may increase (49). Tritiated imipramine binding (3H-IMI) on platelet membranes has been used to assess 5-HT transporter mechanisms (25). There are suggestions that ECT may normalize decreased 3H-IMI platelet binding, but time course and a relationship with clinical outcome are uncertain. Recently, tritiated paroxetine binding has been shown to provide a more specific measure of the high-affinity 5-HT transporter site on platelet membrane, but effects with ECT have yet to be reported. Prolactin response to thyrotropin-releasing hormone (TRH) and to fenfluramine are believed to be serotonergicly mediated. In both cases, there is some evidence that these responses are enhanced following ECT. Overall, this work suggests that ECT enhances serotonergic throughput and function (25, 40, 49). The fact that antidepressant medications and ECS have distinct effects on the 5-HT2 receptor may be relevant to the utility of ECT in medication-resistant patients and its superior efficacy in psychotic depression relative to monotherapy with antidepressant medications.

Antipsychotic effects are exerted by ECT across a range of diagnostic categories (e.g., psychotic depression, mania, schizophrenia), and yet ECT also has antiparkinsonian properties. This would suggest a distinct profile on alterations of dopaminergic function relative to classic antipsychotic medications. Recent microdialysis studies suggest that the concentration of dopamine and its metabolites are markedly increased acutely following a single ECS and that there appear to be increases in basal levels with chronic administration (40). This work also suggests that there is regional specificity in these changes in dopamine levels. A highly consistent finding is that there is increased behavioral responsivity to dopamine agonists following single or chronic ECS. This potentiation of dopamine-mediated behavior is contingent on intact noradrenergic, but not serotonergic, innervation. Effects on the D2 receptor have not been observed. Rather, there is evidence for increased D1 receptor density and increased second-messenger potentiation at this receptor (16). In addition, there is preliminary evidence that behavioral responsivity is enhanced with D1 receptor agonists but not with D2 receptor agonists. In man, growth hormone or prolactin response to apomorphine has been used as a neuroendocrine probe of dopaminergic function. Although there are reports that these responses are enhanced following ECT, there are also negative findings (25). Plasma HVA appears to be unaltered following a course of ECT and CSF HVA has been found to increase shortly following a single treatment. Most reports indicate no persistent effect of ECT on CSF HVA; however, a recent study in medication-free patients reported a marked increase (49). In general, the findings in humans are less consistent than in animals, but similarly suggest enhanced dopaminergic function. This effect appears tied to regionally specific alterations of D1-receptor physiology. Furthermore, there is evidence that the acute release of dopamine and its metabolites is sensitive to the dosage of electrical stimulation and is not contingent on seizure induction. Enhanced dopamine concentrations are not observed with chemically induced seizures, but are unaltered when the seizure discharge with electrical stimulation is blocked by concurrent administration of a barbiturate. If these effects are relevant to the antiparkinsonian properties of ECT, they suggest examination of electrical stimulation and pharmacological combinations that block the production of generalized seizures.

Chronic ECS results in significant but small reductions in muscarinic cholinergic receptor density in cortex and hippocampus and a functional decrease in second messenger response in the hippocampus (40). This is coupled with acute reductions in brain acetylcholine levels and acute increases in choline acetyltransferase and acetylcholinesterase activities shortly following ECS. In humans, there is evidence for increased CSF acetylcholine levels following ECT and spontaneous seizures (25). These findings are compatible with a reduction in central cholinergic function following ECT. Supporting this, chronic ECS has been shown to produce reduced behavioral responsivity to the muscarinic agonists pilocarpine and arecoline. These effects are compatible with a view that ECT exerts antidepressant effects through a reduction of cholinergic supersensitivity. However, reduction of cholinergic tone theoretically should be counterproductive in the treatment of mania and ECT has marked antimanic properties (36). Cholinergic mechanisms are strongly implicated in various aspects of learning and memory. Regionally specific reduction in cholinergic function may be relevant to the cognitive side effects of ECT (30).

In animals and humans electrical induction of seizures results in a subsequent increase in seizure threshold (55). ECS raises the threshold of a variety of convulsant agents that produce seizures through antagonism of g-aminobutyric acid (GABA). Pharmacological specificity obtains, since ECS will not raise the threshold for seizures evoked with a glycine antagonist or a serotonergic agonist (20). This suggests that ECS results in a functional increase in GABAergic activity. Indeed, single and repeated ECS raises GABA concentrations in specific brain regions. Furthermore, ECS, like most antidepressant medications, results in increased density of the GABAB receptor (40). Behavioral response to the GABAB agonist, baclofen, appears to be enhanced following ECS. In turn GABAB receptors modulate the release of other neurotransmitters. For example, chronic administration of the GABA-mimetic agent, progabide, produces many of the same changes in monoamine biochemistry as ECS, including up-regulation of 5-HT2 receptors (25). There is preliminary data that some GABA-mimetics may have antidepressant properties. Complicating interpretation is evidence that GABA synthesis and release are reduced acutely following ECS (40). Several studies have reported reduced plasma and CSF levels of GABA in depressed patients. The one study to examine plasma GABA levels during ECT found acute reductions immediately following seizure elicitation. This effect was unusual in that virtually every other substance in plasma showed no change or an acute surge immediately following ECT.

Chronic ECS increases met-enkephalin and b-endorphin concentrations and synthesis in several brain regions (40). Like models of morphine tolerance, chronic ECS also results in changes in binding to a number of opioid ligands, with suggestions that the d-opioid receptor may be the most affected. Consistent with these data is behavioral evidence that ECS exhibits cross sensitization to acute morphine challenge. In addition, a variety of postictal behavioral phenomena (e.g., antinociception, hyperthermia, hypoventilation, EEG slowing) have been modulated by selective opioid antagonists. There is strong evidence that immediately following ECS one or more endogenous anticonvulsant substances are released in CSF. Antagonist and ultrafiltration studies implicate a large-molecular-weight opioid peptide. Similarly opioid antagonists can block the progressive decrease in the severity and duration of seizures elicited by ECS. Antagonist studies have also modulated the extent of retrograde amnesia following ECS (30). In humans, there are marked acute increases in plasma b-endorphin immunoreactivity immediately following ECT (25). Although it seems likely that ECT results in enhanced function in some aspects of opioid systems, there has been limited research with humans using selective subtype antagonists to determine the involvement of these systems in anticonvulsant, amnesic, or therapeutic effects.

Adenosine has been implicated in the regulation of cerebral excitability. In addition to the increase in seizure threshold, seizure duration shortens during the course of ECT (55). Adenosine antagonists, such as caffeine or theophylline, increase the duration of seizures elicited with ECT. In humans, there is evidence that the effects of caffeine on seizure duration are not associated with a change in seizure threshold, again indicating at least partial independence of these phenomena. Chronic ECS results in a persistent increase in the density of A1 adenosine receptors, whereas effects on A2 adenosine receptors have been inconsistent (40).

The development of in situ hybridization techniques has led to extensive research on ECS effects on gene expression. Electroconvulsive shock results acutely in induction of a variety of protooncogene products, including c-fos, c-jun, jun-B, and zif/268. The regional distribution of these effects and their time course differ qualitatively from the changes associated with stress manipulations and other methods of seizure induction (e.g., caffeine) (40). This approach may offer a novel method for identifying the common anatomic systems activated by seizure induction methods which possess therapeutic properties (ECT, flurothyl, metrazol). Since oncogenes, like c-fos, are regulatory, it would be expected that ECS changes the expression of many proteins. Indeed, there are preliminary findings that ECS results in altered messenger ribonucleic acid (mRNA) levels in a variety of enzymatic, peptidergic, and transmitter systems. These include mRNA coding for the a1- and g2-GABAA receptor subunits in hippocampus and cerebellum, b1-adrenergic receptor in frontal cortex, 5-HT2 receptor in frontal cortex, tyrosine hydroxylase and neuropeptide Y in the locus coeruleus, cerebral ornithine decarboxylase in cortex, somatostatin in the hippocampus, preproenkephalin in hypothalamus and entorhinal cortex, prodynorphin in hypothalamus and hippocampus, peptidylglycine a-amidating monooxygenase in the hippocampus, and preprocholecystokinin, preprotachykinin-A, corticotropin-releasing factor, arginine vasopressin, and heat shock cognate protein. Initial findings with this approach suggest notable differences in how ECS and antidepressant medications modulate the same transmitter system. For example, chronic ECS and imipramine both result in down-regulation of the b1-adrenergic receptor in the frontal cortex. Electroconvulsive shock produces decreased b1-mRNA levels in a time-dependent fashion, parallel to the receptor changes. In contrast, imipramine appears to produce an initial increase in b1 mRNA levels, and decreases are seen only after approximately 3 weeks of treatment.

One class of theories regarding the mechanisms of action of ECT has concentrated on the neuromodulatory functions exerted by endocrine hormones. In general, these theories have emphasized the effectiveness of ECT in treating vegetative or endogenous symptoms, presumed to reflect hypothalamic dysregulation (15). In addition, there has been a view that the generalized seizure in ECT, a fundamental constituent of efficacy, emanates from a centroencephalic pacemaker. Finally, it is evident that a variety of peptides are elevated in plasma immediately following seizure induction in humans (50).

There are limited data supporting these conjectures. Electroconvulsive shock appears to have complex effects on brain concentrations of TRH and on TRH-receptor function, although some have reported consistent increases of TRH concentration following ECS in specific brain regions (40). Early studies suggested that following ECT most depressed patients showed decreased blunting in thyrotropin stimulating hormone (TSH) response to TRH and that manifestation of this change was predictive of long-term remission. More recent work failed to replicate this observation and, if anything, observed increased blunting following ECT (7). A recent preliminary study suggests that L-triiodothyronine (T3) supplementation of ECT may improve clinical response and reduce cognitive side effects (30).

Investigations of the hypothalamic-pituitary-adrenal axis have proved inconsistent. There is preliminary evidence that the CSF level of corticotropin-releasing hormone is increased following ECT. The dexamethasone suppression test (DST) has not been useful in predicting ECT outcome, and it appears that ECT may have opposite short-term effects on the DST (9). Most patients show decreased postdexamethasone cortisol levels following ECT, but an effect of ECT producing abnormal DST responses has also been detected (9). In any case, changes in DST response appear to be independent of clinical outcome. Surprisingly, it was recently observed that plasma levels of exogenously administered dexamethasone typically increase during and following ECT, and this effect was associated with the degree of symptomatic improvement (9).

Some have interpreted the adverse cognitive side effects of ECT as reflecting irreversible structural brain damage (50). The transient nature of most cognitive changes is incompatible with this position, although the permanent gaps in memory for events surrounding the treatment course hypothetically could reflect either functional or structural effects. The evidence from neuropathological investigations has been recently reviewed (8). Human postmortem studies have not linked neuronal cell loss to current ECT practice. Prospective structural brain imaging studies in cohorts of ECT patients have not observed any changes (8). The most critical work in this area involves animal investigation in which more intensive treatment paradigms can be used with more sensitive pathological techniques. Controlled studies using perfusion fixation techniques have failed to observe pathological changes following ECS. Indeed, cell counts in hippocampal fields in animals receiving intensive and chronic ECS have shown no difference from controls, despite the same technique demonstrating cell loss in epilepsy patients with frequent seizures (8). Furthermore, considerable information has accrued about the metabolic and molecular conditions required for cellular death following seizures (22, 60). Conservatively, under conditions of oxygenation, seizures must be continuous for at least 25 to 30 min to result in cell death. This effect is thought to depend on mismatches between metabolic demand and supply in vulnerable neurons and agonist-receptor interactions that trigger dissipative fluxes of sodium and calcium ions across postsynaptic membranes (22, 60). In any case, there is agreement that the minimum conditions for cellular necrosis do not apply to ECT.

The adverse cognitive effects of ECT are stereotyped and pertain largely to discrete aspects of the retention of new information and the recall or recognition of recent events (52, 63). This pattern suggests that ECT has particularly important effects on medial temporal structures long implicated as subserving long-term memory function (50, 63). A large number of theories have been offered to account for the selective amnesic effects of ECT. Among the most attractive is the view that the anterograde and retrograde memory disturbances reflect dysfunctions of consolidation. Months to years after encoding, information may be vulnerable to loss, because the process of consolidation is not complete. Memories that are more remote in time may be least vulnerable, whereas interference with consolidation may result in permanent loss of recently acquired information. We have noted that ECT transiently inhibits protein synthesis and may disrupt the plasticity mechanisms needed for consolidation or maintenance of long-term memories (27, 60).

Relatively little research has been conducted on the neurophysiological correlates of these cognitive effects. As indicated, ECT results in postictal reductions of cortical CBF and cerebral glucose metabolism. Little is known about the persistence of such effects, although reductions have been observed a week following the ECT course (Nobler et al., unpublished data). Regional reductions in CBF and metabolism in many neurological conditions are associated with particular profiles of cognitive deficit. Such correlations have not been examined following ECT. This is particularly important, because the ECT cognitive pattern is suggestive of medial temporal lobe dysfunction; yet the most prominent effects reported in brain-imaging studies have been in the anterior frontal cortex. In contrast, it has been established that the induction of slow-wave EEG activity may persist for weeks following ECT. Several studies have shown significant correlations between the magnitude of this effect and that of amnesic deficits (1, 68). However, this work rarely involved examination of topographic EEG changes.

In animals, ECS is a standard screening procedure for pharmacological compounds that may have beneficial effects on memory. A large variety of agents have been shown to diminish or block ECS amnesic effects, including opioid antagonists, calcium-channel blockers, cholinesterase inhibitors, acetylcholine precursors, corticosteroids, thyroid hormone, vasopressin analogs, somatostatin, melanocyte-stimulating hormone, atrial natriuretic peptide, substance P, nootropics, ergoloid mesylates, and psychostimulants (30). Typically, when the efficacy of such agents has been evaluated in humans, it has been in the context of clinical trials targeting various forms of dementia, often with disappointing results. There have been few attempts to use pharmacological strategies to reduce the cognitive side effects of ECT, despite the wealth of candidate agents and similarities in the cognitive effects of ECS and ECT (30). This approach holds considerable promise for improving this form of treatment. It is evident that distinct pathophysiologies underlie the cognitive disturbance in most dementia conditions and that associated with ECT. Efficacy of pharmacological strategies in one context and not the other would underscore this distinction. On the other hand, repeated failure to discern protective effects in human ECT for compounds that reduce the amnesic effects of ECS in rodents and nonhuman primates would introduce uncertainty about the relevance of particular animal models of amnesia.

Electroconvulsive therapy is a highly effective treatment for episodes of major depression and mania. Its role in the treatment of exacerbations of schizophrenia requires reconsideration. Considerable progress has been made in optimizing ECT technique. This work indicates that technical factors in ECT administration have pronounced impact on both efficacy and cognitive side effects and has contradicted the original view that the generalized seizure provides both the necessary and sufficient conditions for antidepressant effects. The central clinical issue in the use of ECT is the problem of rapid relapse. Research is needed into the efficacy of traditional and alternative continuation medication strategies, as well as the use of continuation ECT, particularly in relation to history of medication resistance. Brain-imaging studies are beginning to outline the functional anatomy of the neural systems associated with antidepressant effects. This work needs to be extended to adverse cognitive consequences. In addition, animal studies have identified a plethora of chronic ECS effects on neurochemical systems. Extension of this work to humans using in vivo imaging techniques and distinguishing between the nonspecific changes associated with seizure induction and those unique to seizures with therapeutic properties offer new approaches to investigation of mechanisms of action.

Preparation of this chapter was supported in part by grants MH35636 and MH47739 from the National Institute of Mental Health.

published 2000