Abstract
Since the initial controlled report of the effectiveness of subconvulsive repetitive transcranial magnetic stimulation (rTMS) by Pascual-Leone in the Lancet in 1996 [1], there has been considerable interest worldwide in the potential clinical use of this technique for the treatment of depression. rTMS is able to stimulate focal areas of brain cortex non-invasively using magnetic fields. Unlike electroconvulsive therapy (ECT), it does not involve a general anaesthetic or seizure.
A possible clinical application was first surmised when several researchers using TMS as an investigative tool in neurology noted incidentally that a number of subjects reported mood changes after TMS stimulation [e.g. 2]. As a result of these fortuitous observations, studies of TMS as a potential treatment for depression were initiated.
One of the current main controversial issues is whether the accumulated evidence for the efficacy of rTMS is sufficiently substantive to justify regulatory approval for widespread clinical usage. This article evaluates the efficacy outcomes reported so far in clinical trials of rTMS, details studies undertaken by our own group, and reviews determinants of the efficacy of rTMS.
Studies of the efficacy of rTMS in depression
There have been a large number of sham-controlled rTMS treatment trials, with over 25 published by early 2006. As there have been relatively small samples in each, we will initially review six published metaanalyses [3–8].
All but one of the meta-analyses has found active rTMS to be statistically superior to sham. However, the major contention has been the clinical significance of the effect. The only meta-analysis to find a negative result applied selection criteria that excluded the majority of trials, leaving little power to detect any significant difference [8].
The earliest meta-analysis [3] only included five controlled studies and excluded the only negative depression trial for rTMS published at the time [9]. Holtzheimer et al. included 12 controlled rTMS depression trials, 11 of which involved rTMS to the left dorsolateral prefrontal cortex and seven of which involved a parallel design [4]. This is a critical distinction, as the blinding of subjects in crossover studies is questionable, given the difference in scalp sensation with active and sham rTMS [10]. They calculated the weighted mean effect sizes to be 0.81–0.89 depending upon which studies were included. These outcomes represent large effect sizes favouring active rTMS Aver sham. However, the authors noted that the studies analysed did not demonstrate a high degree of clinical efficacy, with only 22% of subjects showing response (defined as a reduction of the Hamilton Rating Scale for Depression [HRSD] of at least 50%) with active rTMS, compared with 7.9% of subjects receiving sham treatment.
Burt et al. [5] reported on 16 controlled trials, finding a weighted mean effect size of 0.67. This represents a moderate to large effect size. Although the results show strong statistical support for the efficacy of rTMS, the clinical results were not impressive, with subjects improving (in mean depression rating scale scores) by 23.8% with active treatment and 7.3% with the sham. Kozel and George [6] calculated a mean effect size of 0.53 for 10 controlled depression trials of left prefrontalrTMS.
In a Cochrane review, Martin et al. [7] examined 14 randomized controlled studies. They found that high frequency (<1 Hz) left prefrontal rTMS and low frequency (≤1 Hz) right prefrontal rTMS were statistically superior to sham treatment, but only at one time-point – immediately after the 2 weeks of treatment, with the difference not sustained 2 weeks later – and only for the HRSD. As the Overall difference between active and sham treatment was not large, although significant (a standardized mean difference of −0.35 for high frequency left prefrontal rTMS), they concluded that there was ‘no strong evidence for benefit from using TMS to treat depression, although the small sample sizes do not exclude the possibility of a benefit’.
The only meta-analysis to find a negative result for high frequency left dorsolateral prefrontal rTMS compared with a sham control excluded 13 of 19 eligible trials [8]. The remaining six trials analysed had low power to detect a difference. Overall the meta-analyses indicate that rTMS treatment is superior to a sham control, although a 2-week treatment course provides modest clinical outcomes. Studies published since these meta-analyses were undertaken [11–23] have generally supported the finding of antidepressant effects for rTMS.
Comparison of rTMS and ECT
Several studies have randomized depressed subjects to receive rTMS or ECT. Although these are often quoted as demonstrating comparable efficacy for rTMS and ECT, a careful inspection of the results actually indicates a superior efficacy for ECT [24–27]. In some of these studies the emerging trend in favour of ECT did not reach statistical significance [26], or was only significant on some measures [27]. Of interest, Grunhaus et al. [24] found a clear difference between psychotic and nonpsychotic depression, with rTMS equivalent to ECT only for the latter group.
It is of interest that the mean improvement with rTMS in these studies (40–56% decrease in HRSD scores) is higher than in most of the non-ECT comparator depression treatment trials of rTMS. With no sham comparison in these studies, interpretation of this finding is difficult. These studies have also been criticized for their modest ECT response. In comparison with the effect sizes quoted above for rTMS (0.35–0.89), Burt et al. [5] calculated effect sizes of 2.26 for bilateral ECT and 2.12 for highdose right-unilateral ECT.
Potential determinants of efficacy in rTMS studies
We will now examine some of the possible predictors of response in rTMS trials. Some of these relate to study design – such as the nature of the comparator sham rTMS treatment and the associated issue of the choice of a crossover or parallel treatment design – or the stimulation parameters employed. Other important factors appear to be some of the clinical characteristics of the sample recruited.
What is an appropriate sham for rTMS?
The choice of an appropriate sham form of TMS has been a complex methodological issue. Tilting the active treatment coil to reduce cortical stimulation but preserve the sensation of scalp stimulation (which is important for the simulation of real TMS) may involve low levels of cortical stimulation [10, 28], while forms of sham TMS with negligible stimulation are less convincing and subject blinding may be ineffective, particularly where subjects are not TMS-naive, for example, in studies with a crossover design.
It is apparent that the response rates of the ‘sham’ groups in the early crossover studies were surprisingly low, with either no improvement or given deterioration in this arm in many such trials. This repeated finding suggests some deficiency of the blinding procedures, as analogous pharmacological or ECT trials in treatmentresistant depressed samples invariably demonstrate a considerable placebo response rate, albeit less than in less treatment-refractory samples. This phenomenon of minimal response rates in the ‘sham’ arm surprisingly has also been reported in a number of the more recent ‘positive’ studies which have utilized parallel group designs.
rTMS stimulation parameters
There are many variations in the way rTMS can be given as a clinical treatment, involving choices Aver treatment site, stimulation parameters, and treatment course. Studies to date have differed with respect to the above, with almost no two studies using identical rTMS parameters, except in deliberate attempts at replication.
Stimulation site
High frequency rTMS to the left prefrontal cortex has been used in most trials, a choice influenced by positive early results for this approach [1]. In that seminal report, 17 subjects received five week-long courses of high frequency rTMS (10 Hz) at 90% of motor threshold. Using a crossover design (1 week stimulation, 3 week washout), each patient received treatment at three scalp sites (left dorsolateral prefrontal cortex [DLPFC] – active and sham; right DLPFC – active and sham; and ‘control’ treatment to the scalp vertex). The only active treatment that differed from the sham or vertex stimulation was that administered to the left DLPFC.
Some investigators have trialled low frequency (≤1 Hz) rTMS to the right prefrontal cortex. There are several theoretical reasons for this approach. Low frequency rTMS is relatively safe, with a much lower risk of inducing an accidental seizure [29]. It is more tolerable at a given stimulus intensity and does not require expensive TMS machines capable of high frequency stimulation. Furthermore, motor cortex studies suggest that high and low frequency rTMS have opposite effects on the excitability of neurones in brain cortex [30, 31]. There is considerable evidence from neuropsychological, lesion and imaging studies that the left and right hemispheres have contrasting roles in mood regulation [32, 33]). Therefore, it might be expected that low frequency rTMS to the right prefrontal cortex may be as likely to have antidepressant effects as high frequency rTMS to the left prefrontal cortex.
A number of sham-controlled studies have examined the antidepressant effects of low frequency right prefrontal rTMS in depressed subjects [13, 14, 16,34–36]. In the first study Klein et al. [34] reported promising results with half of the subjects improving by more than 50% in HRSD scores in the active group, compared with a 25% response rate in the sham group. Replication studies using the similar rTMS treatment parameters have reported both negative [13] and positive findings [16, 36]. Other studies using a much higher number of TMS pulses either failed to find a significant difference between sham and active treatment [35] or found significant differences, although the Overall degree of improvement was small [14]. Some of the above studies directly compared low frequency right prefrontal rTMS and high frequency left prefrontal rTMS, but failed to find significant differences between the two approaches [13, 14, 35].
Another approach has been to employ ‘bilateral’ rTMS, that is, rTMS to both left and right prefrontal cortices. In the first study of simultaneous bilateral prefrontal rTMS, Loo et al. tested the hypothesis that the superior efficacy of bilateral ECT Aver unilateral ECT arose from the stimulation of both frontal cortices [11]. Subjects received simultaneous high frequency rTMS to both prefrontal cortices or a sham control. Results were negative, with suggestions that high frequency right prefrontal rTMS may have induced tearfulness in some subjects [11].
Fitzgerald et al. have recently reported on a trial involving ‘sequential bilateral’ rTMS [23]. This comprised three trains of low frequency rTMS to the right prefrontal cortex of 140-second duration at 1 Hz, followed immediately by 15 trains of 5-second duration of high-frequency left-side rTMS at 10 Hz. This treatment was compared with a sham form of rTMS. Aver the first 2 weeks the treatment was more efficacious than the sham therapy, with the authors claiming greater response rates than seen in prior studies. As in the study of Loo et al. [11], there was no ‘unilateral’ arm.
Stimulus frequency
Most studies have given rTMS within the range of 5– 20 Hz, as changes in cortical excitability have been demonstrated with this range of frequency [31]. Preliminary results from rat studies such as our own [37] suggest that higher stimulus frequencies may have greater antidepressant potency. Four sham-controlled studies have directly compared rTMS at different stimulus frequencies with the left prefrontal cortex [21,38–40]. Padberg et al. [38] stimulated at 10 Hz and 0.3 Hz, George et al. [39] and Su et al. [40] used 20 Hz and 5 Hz rTMS and Miniussi et al. [21] compared 17 Hz and 1 Hz rTMS. The two earlier studies found that the lower frequency stimulation yielded superior outcomes, although a statistical difference between the two groups was not demonstrated and the sample sizes (18 and 30 respectively) were too small for firm conclusions. Su et al. [40] found no difference and Miniussi et al. [21] reported better results with 17 Hz than 1 Hz rTMS, although the comparison was only Aver a 1-week treatment period.
Thus, the optimum stimulus frequency for antidepressant efficacy remains unclear and the field awaits large, adequately powered trials directly comparing the effects of a range of treatment frequencies in depressed subjects.
Stimulus frequency and site of stimulation
A number of studies have compared forms of rTMS involving a variety of combinations of stimulus frequencies and stimulation to the left and right prefrontal cortices. In a novel study, Conca et al. [41] assigned 36 subjects to receive either 10 Hz rTMS to the left prefrontal cortex and 1 Hz rTMS to the right prefrontal cortex consecutively in each session (i.e. non-simultaneous bilateral rTMS), 10 Hz and 1 Hz rTMS (alternating trains) to the left prefrontal cortex, or 10 Hz to the left prefrontal cortex (i.e. a control group receiving more typical rTMS treatment). The total number of stimuli was the same for each group. All groups improved but no significant differences were found between them.
Hausmann et al. [42] compared 20 Hz left prefrontal rTMS alone, 20 Hz left prefrontal rTMS followed by 1 Hz right prefrontal rTMS, and a sham control. No differences were found between the groups despite the use of relatively intense rTMS parameters (i.e. long stimulus trains, large number of stimuli). In a comprehensive study, Christyakov et al. [22] randomized depressed subjects to receive placebo medication and rTMS to the left or right prefrontal cortices, at 10 or 3 Hz, or to sham rTMS and clomipramine, Aver a 2-week period. Results clearly favoured left prefrontal rTMS at 3 Hz, although the advantage Aver 10 Hz left prefrontal rTMS could have also resulted from the higher stimulation intensity used for the 3 Hz group.
Stimulus intensity
Most rTMS trials have reported stimulus intensity relative to the subject's resting motor threshold, that is, the lowest stimulus intensity necessary to produce a motor response in a relaxed contralateral muscle when TMS is given Aver the primary motor cortex. Researchers have reported that stimulus intensity is an important factor in determining the induction of long-term potentiation after 7 days of rTMS in rats, with subthreshold rTMS giving the best results [43]. Stimulus intensity may be therefore an important factor in inducing lasting changes in cortical excitability that may be responsible for antidepressant effects. In depressed subjects, two controlled studies compared different intensities of stimulation – 100% and 90% of motor threshold, and a sham condition mathematically modelled as corresponding to 40% [44], and 100%, 80% and sham [45]. Mathematical modelling also demonstrated that higher intensities of stimulation were associated with higher peak current intensities in underlying brain cortex. In these studies, higher stimulation intensity was significantly associated with greater improvement. However, an Overview of all the sham-controlled studies does not support an association between higher intensities and greater response rates, with a large proportion of negative studies using stimulus intensities at the higher end of the range (110–120% motor threshold).
Number of stimuli, length of treatment course
Similarly, the published sham-controlled studies do not support the contention that treatment protocols using a higher total number of TMS stimuli per session have superior therapeutic outcomes. The few studies that reported a 4-week course of rTMS have generally demonstrated continued improvement Aver the third and fourth weeks, although apart from three studies [20, 36, 46], these studies were either not sham-controlled [24–26] or provided continued rTMS in an open extension after an initial sham-controlled phase [9, 14]. In the recent 6-week study of Fitzgerald et al. [23], there was a continued significant difference Aver 4–6 weeks, but interpretation is difficult as only those who responded at the various intervals after 2 weeks were administered ongoing blinded treatment, and drop-out rates in the ‘sham’ group were very high. Given that optimal efficacy is not achieved with other antidepressant treatments (e.g. medications) in a 2-week period, it may be expected that a treatment period of greater than 2 weeks would be necessary for optimal outcomes with rTMS.
Clinical predictors of response to rTMS
There are suggestions that rTMS may be less effective in psychotic depression [24, 25] and in those with depressive episodes of longer duration [17]. Response rates also appear to be reduced in elderly subjects [38, 47]. A possible explanation for this latter finding has been provided by reports such as that of Mosimann et al. [48] who found that in older depressed subjects, an increased distance between the stimulating coil and brain cortex, that is, greater prefrontal atrophy, was associated with a poorer antidepressant response to rTMS. Kozel et al. [49] did not find this association in a younger sample. In these studies greater coil-cortex distance is considered to be a proxy for stimulation at reduced intensities as the intensity of the magnetic field in TMS diminishes rapidly with distance from the stimulating coil [50].
Some studies [e.g. 39] have suggested that those with bipolar depression are more likely to respond than unipolar depressed patients, although such apparent associations are more descriptions than firmly statistically based. Features of melancholia have been variably reported to be positively [14] or negatively associated [23] with response to rTMS. It is also likely that subjects who are more treatment-resistant would be less likely to respond to rTMS, as is the case with other antidepressant treatments [51]. In general terms, this depressive profile is distinct from that associated with a high likelihood of response to ECT, that is, older age with psychotic and melancholic features.
Taking the sample characteristics into account is important when comparing or pooling the results from different studies as they may obscure the apparent relative benefits of different forms of rTMS. For example, it is now widely accepted that the large number of responders to low frequency right prefrontal rTMS in the Klein et al. [34] study is due at least in part to the non-resistant nature of their sample.
Safety
In general, clinical trials have confirmed rTMS to be without significant adverse effects in the treatment of depression. For a review of adverse effects of rTMS in normal subjects and other experimental conditions, see Wassermann [29].
A small number of serious adverse effects have been reported with rTMS treatment in clinical trials. Seizures have occurred in two depressed subjects receiving rTMS [1, 41]. In both cases stimulation parameters were relatively intense and were outside recommended safety guidelines [29], and both patients had commenced medications which probably lowered their seizure thresholds (amitriptyline and haloperidol; venlafaxine) just prior to the incidents. Where rTMS is given within suggested parameter limits and subjects are carefully screened for seizure risk, the risk is very low. Repetitive transcranial magnetic stimulation has also induced mania in a few depressed bipolar patients [40, 52, 53], and hypomania in a subject with unipolar depression [54].
Apart from the above, the majority of depressed subjects receiving rTMS treatment have reported no adverse effects or minor side effects. In most cases these have been pain or discomfort during stimulation (owing to stimulation of scalp nerves and muscles) and headache, often after rTMS. Repetitive transcranial magnetic stimulation has been safely given to a pregnant depressed subject [12], but evidence for its safety in pregnancy is only anecdotal at this stage.
Several studies administered comprehensive batteries of neuropsychological tests before and after a course of rTMS treatment, without finding any significant deficits after rTMS. As practice effects (leading to improved scoring on tests) may have obscured any detrimental effects of rTMS, it is reassuring that no differences were found between subjects receiving a course of active or sham rTMS in controlled studies [11,13–15,17, 18, 36, 38, 46,55–57].
Conclusion
Should TMS be widely clinically available for the treatment of depression? In some countries such as Canada and Israel, rTMS is formally approved by regulatory authorities for such use. In Australia, the Royal Australian and New Zealand College of Psychiatrists Position Statement on TMS [58] recognizes the need for further research into TMS but cautiously allows for its clinical use to treat depression in limited circumstances, with recommended caveats, including that the patient sign a consent form acknowledging that the most efficacious manner of administering rTMS has not yet been established.
In Australia, there has been debate on the advisability of broadening the use of rTMS to clinics beyond research centres, with discussion of associated regulatory issues [59, 60]. Apart from considerations of safety, adequate regulatory safeguards and the limitations of current therapeutic options in treatment-resistant depression, a critical appraisal of the efficacy of rTMS as a treatment for depression is central to the discussion.
As reviewed in this paper, most of the sham-controlled data for the efficacy of rTMS in treating depression comes from studies with a 2-week sham-controlled period. The outcomes suggest clear statistical proof of superiority Aver placebo, but marginal clinical effects. The profile of those more likely to respond is that of younger, non-psychotic and less treatment-refractory patients – a group distinct from those who tend to benefit from ECT. There is clearly a need for longer controlled trials of 4–6 weeks in duration to clarify if longer courses enhance the likelihood of improvement.
Furthermore, the most appropriate role for rTMS as a treatment in depression, whether it be a stand-alone treatment, add-on to other treatments (such as antidepressants) for synergistic effects [20, 42, 52, 61], or as an alternative means of inducing seizures [62] is yet to be determined.
It is our considered opinion that widespread prescription of rTMS for depression would be premature on the basis of current knowledge. However, as we write, large multicentre US-based commercial and governmental (NIMH-funded) trials are well underway, with some Australian centres involved in the former. These trials should provide sufficient statistical power to clarify the efficacy and safety of rTMS in sample sizes akin to those required by national regulatory authorities for approving antidepressant medications.
Footnotes
Acknowledgements
Dr Loo is a part-time Senior Research Fellow supported by a National Health and Medical Council Program Grant no. 2223208. The authors thank Kate Manollaras for assistance in manuscript preparation. We are also grateful for funding support from the Black Dog Institute Foundation and associated private benefactors (Peter Joseph and Bill Loewenthal).
This paper was presented at the Joint CINP/ASPR meeting, held in Brisbane, December 2005.
Some of the material in this article has also appeared in other publications by the authors including Loo & Mitchell 2005, Journal of Affective Disorders 88: 255–267.