Few medical treatments have attracted as much public and medial attention as electroconvulsive therapy (ECT). In popular movies and politically oriented journalism, ECT is often pictured as repressive and inhumane. Promoted by antipsychiatry movements, such images have had an impact on treatment practice and have contributed to restrictions in the use of ECT in many countries. In Scandinavian and many other countries, the treatment is widely accepted as indispensable in the treatment for deep depression and other severe mental disorders. In spite of an unrivaled and scientifically documented effectiveness, cognitive side-effects are a cause of controversy. Memory impairment after the treatment is also a major concern for the USA Food and Drug Administration (FDA), which is to decide whether ECT shall remain a class III (high risk) medical equipment, requiring premarket approval, or be reclassified to class II (intermediate risk, without such requirement). Psychiatrists and anesthesiologists (with direct experience of ECT) are in favor of reclassification, whereas neurologists, psychologists, biostatisticians, and public representatives advise retaining class III status (1). Memory impairment and the risk of inducing cerebral lesion are also focused by the founder of the lay anti-ECT group The Committee for Truth in Psychiatry (2) and her successor (3), who both experienced vast erasement of memories following ECT. Contrasting to the emphasis on undesired side-effects is the clinical experience that ECT can induce transition of severely melancholic and suicidal patients into normally functioning and dignified fellow human beings, often when all other treatments have failed. With the aim to attain a reasonable balance between effectiveness and safety, we present the modern base of evidence for benefits and risks of ECT. Similar surveys have been published in USA (4, 5). Over a history of more than 70 years, the indications for ECT have been crystallized into psychotic depression, deep depression with risk of suicide, and some other mental disorders with a high risk of lethal outcome (delirious mania, catatonic stupor, postpartum psychosis, cycloid psychosis, lethal catatonia, and neuroleptic malignant syndrome). In the major indication psychotic depression (with delusions and/or hallucinations), the remission rate is 92–95 per cent, which may be compared with 55–84 per cent in non-psychotic melancholic depression (6, 7). Imipramine in a daily dose of 200–350 mg caused remission in only 40 per cent of patients with psychotic depression, whereas 83 per cent of those not responding remitted after ECT (8). Addition of antipsychotic to antidepressive drugs only insignificantly augments the effect (9). Besides, the combination of drugs involves risk of weight increase and extrapyramidal side-effects, especially in elderly persons. The American Psychiatric Association recommends ECT as treatment of choice in psychotic depression (10). Also in non-psychotic melancholic depression, ECT is superior to drugs and is usually offered when their effect has been insufficient. Meta-analyses show that ECT is more efficacious than simulated ECT and also more efficacious than antidepressive drugs (11, 12). The imminent risk of suicide in depression decreases faster with ECT than drugs and gives cause for a more prominent position of ECT in the treatment for suicidal depression (13). Outside the traditional indications, the effect of ECT is unspecific and short-lived and side-effects run the risk of prevailing. After successful ECT, the risk of relapse is more than 80 per cent within a year, making continued stabilizing treatment necessary (14). The usual means are antidepressants, lithium, and antiepileptic drugs. Continuation ECT with successively increasing intervals after the acute series is an alternative option to prevent relapses and recurrences. In American studies, a combination of nortriptylin and lithium lead to relapse during 6 months in 39 and 32 per cent, respectively, to be compared with 84 per cent with placebo (14) and 37 per cent with ten stabilizing ECT treatments at increasing intervals (15). Subjective memory and performance in several memory tests improved during the stabilizing treatment, and there was no cognitive difference between the treatments (16). Earlier studies of continuation treatment with imipramin and paroxetin have reduced the relapse rate to 20–30 per cent during 6 months (17-19), but the results have not been replicated. If continued ECT and drugs are combined, the risk of relapse may be further reduced. In an open study, seven per cent relapsed during 2 years to be compared with 52 per cent with drugs only. After 5 years, the relapse rates were 27 and 82 per cent respectively. The time to relapse was also prolonged (20). A naturalistic study of 41 patients who received drugs (antidepressive, antiepileptic, neuroleptic, and/or lithium) together with continuation ECT for 3 years showed a 75 per cent decrease in hospital care compared with 3 years before (21). Independent of the method, a condition for successful continuation treatment is an unambiguous effect of the acute series of ECT. If an unspecific anxiolytic effect of ECT is misinterpreted as a genuine antidepressive action, there is risk that too many treatments are given without benefit. Transient memory disturbance is an inevitable side-effect of ECT that does not contribute to the antidepressive effect. Inability to remember comprises particularly the period of treatment but often also before and afterward. Impaired memory of what happened before the treatment – retrograde amnesia – includes both personal and public events. There is a time gradient with memory of events occurring closer to ECT being more affected, especially within 6 months before the treatment. Impaired memory of events after the treatment – anterograde amnesia – is owing to deficient consolidation of new information. The more treatments that are given and the more densely spaced they are, the more obvious and protracted is the memory impairment. Objective memory loss is usually shorter than subjective loss, most studies indicating return to the pre-ECT level or improved memory function within 2–6 months (22). However, a naturalistic study using obsolete bitemporal sine wave stimulation showed impairment of personal memory recall 6 months after ECT (23). A meta-analysis of 2981 patients showed that cognitive side-effects of ECT mainly comprised the first 3 days after the treatment and that all cognitive functions improved compared with pretreatment conditions, irrespective of stimulus parameters (Fig. 1) (24). Short (0–3 days) and long-term (>15 days) cognitive effects of electroconvulsive therapy (24). Computer tomography and magnetic resonance imaging before and after a series of treatments have not disclosed any lesions in the brain (25, 26). A persistent gap of memory after a series of ECT is not only caused by the treatment but also by the mental disorder that was the reason for treatment. Serious symptoms and chaotic live conditions that interfere with attention and concentration cause a gap of memory for the time of illness, whether ECT has been given are not. Over time, the memory gap diminishes with the most distant events returning first, but the time of treatment often becomes a permanent gap. In exceptional cases, more extensive and long-lasting memory impairment is experienced. The background is often complex, and court proceedings in USA have not displayed an unequivocal relationship with ECT and damages have not been imposed (27). It belongs to the conditions of human life that personal memories gradually get lost if they are not regularly recalled, retold, or rehearsed. No test of retrograde amnesia used in ECT research is standardized for normal forgetfulness. Experience shows that patients who get an obvious relief of their suffering from ECT seldom complain of memory impairment. Compared to the freedom of a painful depression, a possible impairment of memory is experienced as a minor inconvenience rather than a handicap. Some even experience their memory as improved (28). If, on the contrary, the antidepressive effect has been doubtful, memory problems may prevail. This illustrates that ECT should mainly be offered to patients who are anticipated to benefit substantially from the treatment. In addition, if the treatment is not discontinued when signs of improvement fail to appear, the only effect may be memory impairment that adds to the original symptoms. The same may happen if the treatment is given on appropriate indication but discontinued before a stable remission has been attained. Severe depression may have a toxic effect on the brain, manifested as decreased volume of hippocampus. The longer the duration of depressive episodes has been, the more marked is the decrease in hippocampal volume. Because the severity of depression is related to the blood level of cortisol, which normalizes when the depression remits, the decrease in hippocampus is probably caused by long-lasting or periodic increase in the level of cortisol in the blood (29-31). When the hormonal balance is restored after ECT, the volume of hippocampus increases (32). In rats, electrically induced seizures stimulate the formation of new neurons (33), glia cells (34), and blood vessels (35). In this animal model of ECT, the formation of new neurons is much greater than of antidepressive drugs. Because there is new formation of all structures, it is unlikely that the glia increase is a consequence of loss of neurons. Microglia controls the formation and development of new neurons. A hypothesis has even been launched that depression originates in glia cells (36). It is likely that ECT exerts its action by regulating the hypothalamic control of the neuroendocrine system, displayed as normalized cortisol level in the blood (37). The importance of the hippocampal increase is uncertain. New cells may be formed without entailing an antidepressive effect, and drugs may have an antidepressive effect without the formation of new cells (38, 39). With ECT, severe depression usually mitigates already after the first treatment. As a rule, twelve treatments at most are needed every second or third day. In older studies of endogenous depression (corresponding to melancholic depression), no more than six to seven treatments on average were required for clinical remission (40, 41). A multicenter study by the American CORE group on the same type of patients showed remission in 64 per cent after an average of 7.3 treatments (15). During recent years, there has been a tendency toward longer acute series and lower rates of remission than before. This is cause for a critical scrutiny of treatment praxis. In order to reduce memory impairment, sine waves or modified sine waves have been substituted with brief or ultrabrief pulses and bitemporal electrode positions have been replaced with right unilateral stimulation. This ambition to limit anticipated side-effects instead of optimizing therapeutic efficacy may have reached an impasse where submaximal seizures are induced with lower antidepressive effect. In Scandinavia and many other European countries, the Siemens Konvulsator has been the device of choice for routine use and research. It delivers unidirectional 5 ms pulses at 50 Hz continuously or in groups of four pulses, reducing the frequency to 25 Hz. The maximum charge is substantially higher than with modern devices, but the low pulse frequency and long time of stimulation made it possible to adopt the dose of electricity individually by discontinuing the stimulation at the start of the tonic phase of the seizure. Modern devices deliver bipolar pulses with amperage up to 0.9 A for 8 s at most. The maximal frequency is 120 Hz, implying up to 240 pulses per second. The high frequency of pulses results in a tonic contraction of musculature, which makes it impossible to discern the start of the tonic phase of the seizure and then discontinue the stimulation. Instead, the dose of electricity must be estimated in advance. The recommendations of manufacturers for the first treatment based on age and sex are only partially helpful. At best, they serve as a rough guidance making it necessary to secure that a fully generalized grand mal seizure has been induced. The motor convulsions should have a tonic-clonic course with successively decreasing frequency and cessation starting in distal parts of the body. The electroencephalographic (EEG) pattern should display initially increasing amplitude (recruitment of neurons), a period of spikes with high frequency and amplitude, transition into a spike-and-wave pattern and a distinct end followed by postictal depression. The pulse rate should increase markedly during the seizure, as well as the blood pressure immediately after the end of the seizure. The seizure should be followed by postictal confusion on awakening. The total dose of electricity (charge) in milliCoulomb (amperage × time) is determined by amperage, frequency of pulses, pulse width, and total time of stimulation. The relative importance of these parameters is still insufficiently understood, but in the interval, 30–60 Hz frequency proved unimportant (42). Stimulation for 2 s compared to 1 s with the same dose of electricity was more efficient (43). The studies offer imperfect guidance, but there is some knowledge of the importance of electrode placement, dose of electricity, and pulse width. When replacing the original bitemporal stimulation with unilateral stimulation of the non-dominant hemisphere (usually the right), the d′Elia positions are usually preferred (41). Scandinavian studies with the Siemens Konvulsator have shown that unilateral stimulation results in less memory impairment but the same antidepressive effect as bitemporal stimulation (41, 44, 45). However, these circumstances do not apply to other widths, amplitudes, and frequencies of pulses that are delivered by modern devices. On the whole, it is technically more difficult to induce maximal seizures with unilateral than with bitemporal stimulation. Barely suprathreshold stimulation, brief or ultrabrief pulses, anticonvulsant drugs, and too deep a narcosis may interact in an unfortunate way to prevent the spread of the seizure from a unilateral focus. Submaximal seizures occur more often with unilateral than with bitemporal stimulation (46) and are the probable cause of lower efficiency of unilateral ECT (47, 48). Because the therapeutic effect of ECT depends on grand mal seizure activity, it may be argued that the mode of induction is irrelevant. However, the electrical stimulation is important in two ways: excessive stimulation causes more memory impairment without increasing the therapeutic effect, whereas insufficient stimulation may induce submaximal seizures with lower therapeutic effect. Therefore, a dose of electricity should be delivered that induces a fully generalized grand mal seizure, neither more nor less. High-dose ECT is more effective than low-dose ECT but tends to cause more memory impairment (11, 49). American investigators often titrate the seizure threshold at the first treatment – i.e., the lowest amount of electricity that induces any kind of epileptic activity. Titration shows that the therapeutic effect of ECT varies with the dose of electricity in relation to the seizure threshold, especially with unilateral stimulation. A condition for a satisfactory effect of unilateral treatment is that the dose of electricity is several multiples of the seizure threshold. When the unilateral stimulation was only 1.5 times the seizure threshold as many as 10.5 treatments were needed on average to attain remission in 55 per cent of the patients, although, owing to insecure efficiency, bitemporal stimulation was applied after the 5–8 first treatments (14). When the unilateral stimulation was increased to six times the seizure threshold and compared with bitemporal stimulation 1.5 times the seizure threshold, there was still fewer remissions (55 vs. 64 per cent) in spite of somewhat more treatments (5.9 vs. 5.5). Besides, the effect of unilateral stimulation was slower than of bitemporal stimulation. There were no differences in the impact on memory and executive functions. Thus, unilateral stimulation had no cognitive advantage when it reached a level that guaranteed the induction of seizures with nearly as good antidepressive effect as after bitemporal stimulation (50). A good antidepressive effect with high rates of remission has been obtained with 5 ms pulses in Scandinavia (40–43) and 1 ms in American studies (14–16, 48). Also, 0.5 ms pulses have been recommended because they induce seizures with higher heart rate than 1.0 ms pulses, possibly indicating stronger and more generalized convulsive activity. However, the only comparative study is small and does not allow any conclusions on antidepressive effect (51). A further reduction to ultrabrief stimuli (0.1–0.3 ms) often generates submaximal seizures with impaired antidepressive effect on bitemporal stimulation (52, 53). When conventional 1.0 ms pulses were substituted with 0.3 ms pulses in unilateral stimulation, the treatment series were prolonged from nine to 12 treatments on average, and still as few as 13 per cent attained full remission (54). Ultrabrief pulses (0.3–0.37 ms) in unilateral stimulation six times seizure threshold required more treatments than the same pulse width with bitemporal stimulation 2.5 times seizure threshold, which indicates that the combination of ultrabrief pulses and unilateral stimulation is particularly inexpedient (55). It is true that ultrabrief pulses make it possible to induce seizures with lower dose of electricity, but in spite of similar duration and seemingly the same pattern as maximal seizures, EEG displays an atypical course and termination, the increase in heart rate and blood pressure is only moderate and the patient regains consciousness soon after the seizure and is not confused to the same degree as after grand mal seizures. Thus, the advantage of less memory impairment with ultrabrief pulses is often outweighed by lower antidepressive effect. However, contradictory results have been reported from a comparison of brief pulses (1.5 ms) and ultrabrief pulses (0.3 ms) applied unilaterally or bitemporally (56). Whereas, as expected, retrograde amnesia was less with both ultrabrief pulses and unilateral stimulation, ultrabrief unilateral ECT resulted in remission in as many as 73 per cent, brief unilateral ECT 59 per cent, brief bitemporal ECT 65 per cent, and ultrabrief bitemporal ECT only 35 per cent. The EEG-recorded seizure duration was similar in the groups, 50–60 s, as was the average number of treatments, 6.2–8.9. The EEG seizure pattern was not described. The poor outcome in the ultrabrief bitemporal group may have been owing to a too low absolute dose of stimulation, whereas the exceptionally good outcome in the ultrabrief unilateral group may be explained by a dose of electricity than was sufficient to induce grand mal seizures. A higher amount of patients with psychotic depression may also have contributed to the good outcome in that group. Deep narcosis makes it difficult to evoke maximal seizures owing to the inhibition of the spread of paroxysmal activity through the brain. During recent years, brietal and tiopental have been replaced with propofol in many places, leading to shorter seizure durations and lower antidepressive effect (57). Combination of deep narcosis, ultrabrief pulses, and unilateral stimulation requires markedly increased dose of stimulation, often far beyond the capacity of modern devices, if maximal seizures are to be induced in the first place. The great supply of drugs has increased the risk of inappropriate polypharmacy together with ECT. Bensodiazepines have an anticonvulsive effect that may decrease the antidepressive effect of ECT, particularly with unilateral stimulation (58). If possible, bensodiazepines should be phased out before ECT, as long as their half-lives motivate. If continued bensodiazepine treatment is inevitable, a smaller dose of the narcotic drug or the antagonist flumazenil (Lanexat) may be given in conjunction with the narcosis. For the same reason, mood-stabilizing antiepileptic drugs should not be combined with ECT and ongoing medication should be reduced or discontinued. Even if there are reports on increased confusion and memory impairment when lithium is given together with ECT, the combination usually has no negative effects (59, 60). Addition of tryptophan adds only insignificantly to the effect of ECT and has an adverse effect on memory (61, 62). Whereas treatment with nortriptylin has been maintained to increase the antidepressive effect of ECT and reduce the memory impairment, concomitant treatment with venlafaxin only marginally improves the antidepressive effect and impairs memory function (63). Because the antidepressive effect of ECT is so strong, the addition of antidepressive drugs is of minor practical value for the potentiation of ECT, but should be reserved for continuation treatment. In order to obtain an optimal risk/benefit ratio, it is advisable to restrict the use of ECT to severe mental disorders where there are no comparable alternatives, above all psychotic and life-threatening depression, but also delirious mania, catatonic stupor, postpartum psychosis, cycloid psychosis, lethal catatonia, and neuroleptic malignant syndrome. Because in such cases drug treatment is not enough to relieve suffering, ECT should be the treatment of choice. When less severe depression proves resistant to drug treatment, ECT may be offered as second choice. Widening the indications to mental disorders, where ECT has no obvious therapeutic advantage to other treatments but may give rise to memory complaints, runs the risk of undermining the reputation of the treatment. Continuation ECT is an alternative option to stabilizing drug treatment, and the combination of continued ECT and drugs may augment the preventive effect on relapses and recurrences even more. Transient memory impairment is an unavoidable side-effect of ECT that must be weighed against the benefit of the treatment and the risk of suicide if the depression is inadequately treated. Restricting the use of ECT to severe mental disorders conveys an optimal balance between the therapeutic effect and memory impairment. Transient memory impairment is usually experienced as a reasonable price for having obtained relief of severe suffering. The aim in ECT is to combine maximal therapeutic effect with minimal memory impairment. Choosing bitemporal stimulation in the treatment for psychotic and life-threatening depression implies an ethically reasonable prioritization of maximal and fast symptom reduction above minimal memory impairment (64). The opposite priority by choosing unilateral stimulation in less urgent cases is defensible provided that grand mal seizures are induced. More treatments may be needed that may cause the same degree of memory impairment as after shorter series with bitemporal stimulation. The ambition to minimize memory impairment is commendable, but ultrabrief pulses and unilateral stimulation may induce submaximal seizures with decreased therapeutic effect. The optimal pulse width remains to be ascertained but present experience shows that 1.0 ms pulses have satisfactory effect with bitemporal and, with increased dose of electricity, unilateral stimulation. In most studies, pulse widths below 0.5 ms have displayed a weaker antidepressive effect both with bitemporal and unilateral stimulation. However, because the dose of electricity necessary for the induction of grand mal seizures is a function of both the amplitude, width and frequency of pulses, as well as the total duration of the stimulation, pulses below 0.5 ms may be efficient, provided that other stimulus parameters are modified. An obvious possibility is to increase the amperage, but the restricted power of modern devices is a limiting factor, especially in the treatment for elderly patients. The appropriate balance between the electrical parameters should be the object of continued research in order to attain an optimization of the risk/benefit ratio of ECT.