Antidepressant outcomes of high-frequency repetitive transcranial magnetic stimulation (rTMS) with F8-coil and deep transcranial magnetic stimulation (DTMS) with H1-coil in major depression: a systematic review and meta-analysis

Non-invasive brain stimulation methods are established as viable treatment options for major depressive disorder (MDD) resistant to pharmacotherapy. One of such most thoroughly investigated methods is the high-frequency repetitive transcranial magnetic stimulation (rTMS) of the left dorsolateral prefrontal cortex (DLPFC) commonly applied with a figure-of-eight (F8) coil. This treatment aims to ameliorate the hypoactivity of the DLPFC characteristic of MDD and has acute moderate to large antidepressant effects as well as acceptable tolerability [1‐4]. Since 2008 rTMS has been approved for treatment-resistant unipolar MDD by the U.S. Food and Drug Administration (FDA).

One alternative to rTMS with the F8-coil is the deep transcranial magnetic stimulation (DTMS) with the Hesed-coil (H-coil). The H-coil was designed to induce the electrical fields at different locations around the surface of the head that have a common direction and presumably summate in the deep neural areas [5]. While there are several types of H-coils, the H1-coil is comparable to the F8-coil because it stimulates mostly the left DLPFC [6]. However, relative to the focal stimulation achieved with the F8-coil, the H1-coil is less focal [6]. A consistent reduction of symptoms in some substance use disorders indirectly suggests that the H-coils may target deeper reward pathways [7] although the issue of depth is not resolved yet. The literature so far suggests that the H1-coil produces consistent short-term antidepressant effects in MDD [8, 9] and is FDA-approved for treatment-resistant unipolar MDD since 2013.

Although rTMS with F8-coil and DTMS with H1-coil are promising treatments for unipolar MDD, the head-to-head comparisons in efficacy of both methods are lacking so far. A recent network meta-analysis of currently available non-invasive brain stimulation methods concluded that the high-frequency rTMS with F8-coil over the left DLPFC is among the most efficacious techniques, whereas DTMS with H1-coil is not more effective than sham [10]. However, since the analysis focused on the double-blind, randomised-controlled trials (RCT) with inactive sham groups, it compared only one available RCT using DTMS with H1-coil [9] relative to over 50 RCTs using rTMS with F8-coil and a plethora of different stimulation paradigms [10]. Consequently, a comparison in efficacy of both coils might have been affected by the imbalance in the volume of available evidence. Furthermore, there are differences in the FDA-approved protocols for the clinical application of the two devices. Inspection of the stimulation parameters in 54 RCTs with MDD patients [3] revealed that rTMS with F8-coil was most often applied with the intensity of 80–120% of the resting motor threshold (MT) and either 10 Hz for at least 10 daily sessions or 20 Hz in 10 sessions. In contrast, DTMS with H1-coil was always applied with the intensity of 120% MT and 18–20 Hz over 20 daily sessions in 10, mostly open-label, studies with MDD patients [8]. These differences in the stimulation protocols prevent any direct comparisons of the FDA-approved clinical paradigms of both coils in the treatment of MDD.

Similar to the network meta-analysis [10] the aim of the current study is to systematically assess and compare the antidepressant outcomes of rTMS with F8-coil and DTMS with H1-coil in MDD. However, since only one RCT was conducted with the DTMS method in MDD to date [9], we focus on the following parameters rather than study designs:

Although the protocol for this review was not published a priori, we have selected the above parameters a priori based on our knowledge of the stimulation designs used in this field and the statistical approach used in our previous meta-analyses of rTMS and DTMS studies [3, 8, 13]. The approach of focusing on a set of homogeneous parameters, rather than on the FDA-approved clinical paradigms or RCTs only, is suited to assess the antidepressant outcomes of the two coils by controlling for as many potential confounding factors as possible. Since we attempt to assess the effects of active stimulation only in studies with any designs (including uncontrolled, open-label studies), our analytical approach is not designed to quantify the clinical efficacy of either coil.

The current focus of research in the field of the non-invasive brain stimulation is concerned with identifying predictors of response [34] to determine which variants of these methods (conventional F8-coils, H1-coils, or other systems) are most effective in the treatment of MDD [10]. Although F8-coils and H1-coils are FDA-approved for treatment-resistant MDD, surprisingly little is still known about the most optimal stimulation protocols required for best acute and longer-term efficacy of these methods. Although not designed to assess efficacy, the current study provides the first approach to systematically assess and compare the antidepressant outcomes of rTMS with F8-coil and DTMS with H1-coil in unipolar MDD in studies matched on stimulation frequency.

The overall assessment of antidepressant outcomes with either coil (F8-coil and H1-coil) highlights the following main findings. First, the primary outcome (depression severity) was alleviated following the high frequency (18–20 Hz) stimulation with either coil (F8-coil and H1-coil) already after 10 daily sessions relative to baseline in unipolar MDD. However, other meta-analyses have shown that depression severity was reduced even further at session 20 relative to session 10 in rTMS studies with F8-coils [35] or in DTMS studies with H1-coils [36]. Similar, response and remission rates are relatively low after 10 sessions of rTMS or DTMS (29 and 15%, respectively) in the current analysis. Both rates were shown to increase after 20 relative to 10 sessions of DTMS [36]. Therefore, at least 20 daily stimulation sessions may be required for clinically (rather than statistically) meaningful effects of rTMS or DTMS in the acute treatment of unipolar MDD. The open-label extensions to the largest RCTs in the field [2, 9] have shown that some patients experience a delayed response and require 4–6 weeks of stimulation before showing adequate response to treatment with F8-coil or H1-coil [37, 38]. The downside of prolonging treatment is the issue of higher costs and inconvenience of treatment leading to higher dropout rates. One alternative approach in the field aiming to reduce these problems is to provide an accelerated treatment with more than one daily session to induce efficacy faster [39]. However, such protocols still need to be tested in larger RCTs and compared to the single-sessions per day using head-to-head designs.

Second, the primary outcome (depression severity) did not depend on the study design (RCT vs. open-label) nor the therapy type (add-on to antidepressants vs. monotherapy). However, our results suggest that the reduction in depression severity was lower in studies with older patients, as also reported by others [40]. This result could be due to reduced plasticity, connectivity, and the motor threshold in older patients. The effect of age could also depend on the length or the severity of illness that may be higher in older patients. Therefore, future research needs to devise protocols tailored towards the needs of older patients and/or patients with higher illness severity. For instance, older patient groups may benefit most from high intensity stimulation with a large number of pulses due to lower capacity for plasticity as the brain ages [41].

Third, the secondary outcomes (response and remission rates) after the stimulation with either coil (F8-coil and H1-coil) were low. Since these outcomes are not controlled for inactive sham, it is unclear if they are comparable to response and remission rates in RCTs conducted with either coil. The secondary outcomes tended to increase in the open-label studies relative to the blinded RCTs and in studies with patients receiving concurrent antidepressants relative to the monotherapy. These results probably reflect real-world outcomes whereby patients in the clinical practice know what treatment they receive and for ethical reasons are kept on medication even if they do not adequately respond to such pharmacotherapy [8, 42]. Although the current response and remission rates following the active stimulation with either coil are not controlled for sham, these effects cannot be explained by placebo/expectation alone because active stimulation produced better efficacy than inactive sham in several RCTs with the F8-coil [1] and the H1-coil [9]. In general, the efficacy of both methods (F8-coil and H1-coil) with highly variable stimulation protocols is comparable in magnitude based on the moderate acute effects observed in the double-blind, sham-controlled RCTs with MDD patients [1, 3, 9, 43]. In the same vein, the current response and remission rates cannot be attributed to antidepressants alone given that patients in most studies were pharmacoresistant and efficacy was demonstrated in RCTs with F8-coil and H1-coil applied as the monotherapy for MDD [3, 9]. It has been suggested that brain stimulation may enhance or precede subsequent effects of antidepressants for some patients, for instance by rendering neural pathways more susceptible to drug-induced functional plasticity [43].

The comparison of rTMS studies with F8-coil and DTMS studies with H1-coil suggests that there might be some differences in the antidepressant outcomes of these two methods. Our results indicate that the open-label H1-coil studies demonstrated a consistently larger reduction in the primary outcome (depression severity) relative to the F8-coil in all studies. This pattern was also observed when only comparing H-coil and F8-coil studies in open-label designs and in studies with patients concurrently receiving antidepressants, and was thus irrespective of the study design (RCT and open-label) and the therapy type (add-on to antidepressants or monotherapy). In contrast, any differences in the secondary outcomes (response and remission rates) were less consistent. The remission rates tended to favour the F8-coil relative to the H1-coil in all studies, while all other comparisons, including remission and response rates were not statistically significant. The visual inspection of forest plots (Fig. 4) also did not reveal any consistent trends towards differences in the secondary outcomes between both coils. It is unclear if the differences in the primary outcome between the F8-coil and the H1-coil are meaningful (clinically-relevant) or if they are secondary to various factors not controlled for in our study. For example, the H1-coil may reduce depression severity faster than the F8-coil (already after 10 daily stimulation sessions) while the comparison of the same outcome at a later session (20 or later) could reveal no differences between both coils. Such a comparison was not conducted here since most of the rTMS studies did not continue for longer than 10 sessions with the extra-high frequency of 18–20 Hz. The current results suggest that even a large reduction in depression scores may be inadequate to classify the patients as responders or remitters in case their baseline depression severity is high and/or if response and remission require stimulation for more than 10 daily sessions. Thus, not the coils themselves but rather factors, such as time of assessment, baseline depression severity as well as definitions of response and remission based on different versions of HDRS might have contributed to the inconsistent secondary outcome scores observed in this analysis.

It is not surprising that the non-focal stimulation with the H1-coil produces larger reduction in depression severity than the more focal F8-coil, particularly since the target region is difficult to locate in the rTMS studies with the F8-coil [44]. The fact that the more widespread stimulation is likely to target the appropriate region may in itself be an advantage of the H1-coil. In addition, DTMS with H1-coil may also directly influence the activity of deeper, subcortical regions in the emotion regulation network [45]. However, to date it is still being debated whether DTMS is indeed as deep as originally proposed [46] or whether it induces its effects mainly by stimulating a less focal cortical surface relative to a focal stimulation delivered with the F8-coil [47]. A recent mathematical model suggests that relative to 2.2 cm with the F8-coil, the H1-coil can reach up to 6.4 cm while retaining 50% of its maximum electrical field strength [45]. Future investigations by independent groups should corroborate these findings. Finally, the stimulation intensity as well as other parameters not controlled for in this study might have contributed to the greater reduction in depression severity in studies with H1-coil relative to studies with F8-coil. In general, the DTMS studies with H1-coil utilised the high intensity protocols (120% MT, 1680–1980 stimuli/session) relative to the lower intensity protocols (80–110% MT, 800–2000 stimuli/session) in the rTMS studies with F8-coil. However, intensity alone was not related to depression severity in the rTMS studies meaning that a combination of parameters may be required for the most optimal antidepressant outcomes. Since two meta-analyses showed that lower number of stimuli/session (1200–1500) produced greater reduction in depression severity in rTMS studies with F8-coil [3, 35], not intensity alone but rather the combination of frequency-intensity-stimuli may be predictive of antidepressant outcomes. One caveat of the high frequency/intensity of stimulation is the trend towards higher dropout rates in DTMS relative to rTMS studies and the increased risk of seizures with H1-coils due to the larger volume of stimulated tissue [48]. Indeed, two seizures were reported in the DTMS studies with H1-coil and none in the rTMS studies with F8-coil. More research is also required to investigate if the length of the inter-train-interval could also be a predictive factor for the antidepressant outcomes of the F8-coil and the H1-coil.

The systematic assessment of the literature shows that the rTMS studies with F8-coil apply highly variable stimulation protocols to highly heterogeneous groups of patients since 1997 [3]. Although the stimulation protocols are more consistent in DTMS studies with the H1-coil, our analysis did not identify predictors of antidepressant outcomes apart from patient age that may be confounded by duration and severity of illness. This may be a result of a low statistical power in a meta-analysis with data from only 19 studies. However, it is also in line with previous studies that did not consistently identify predictors of response to rTMS or DTMS using primary and secondary analyses [3, 37, 49, 50]. The use of multivariate regression models that probe the effect of one predictor on multiple outcomes while holding all other factors constant is required once an adequate volume of data exist.

It remains unclear for how long the antidepressant outcomes of the non-invasive brain stimulation last in MDD [51]. Given a moderate efficacy to any kind of non-invasive brain stimulation [52], both rTMS and DTMS are viable alternatives for patients resistant to pharmacotherapy in the short-term. Findings regarding the durability of the antidepressant outcomes of both DTMS and rTMS indicate that such effects may last up to six months to a year in some cases [51, 53]. In particular, regular maintenance treatments may prolong the acute antidepressant outcomes and prevent relapse [4, 20]. However, these findings also highlight substantial variability among patients. Maintenance may have to be based on individual patients’ needs as to optimise cost-effectiveness for prevention of relapse.

According to discussions at the 2nd European Conference on Brain Stimulation in Psychiatry (ECBSP, October 2017, Munich, Germany), the important focus in the field of the non-invasive brain stimulation is the individualisation of treatment. However, given the vast options regarding the parameters of stimulation, identifying an optimal protocol may be akin to finding the proverbial needle in the haystack. One way of constraining the number of possibilities is by understanding the factors that influence neuronal communication and in turn plasticity [54, 55]. In fact, such an approach has been used to guide the design of various TMS methods in the past, such as theta burst stimulation [56]. It is also currently being explored to tailor TMS methods in a way that may either interfere with the abnormal or boost the beneficial cortical oscillations in a frequency-specific manner [57, 58]. Such applications may be tailored to an individual’s endogenous neuronal activity but do not have a substantial evidence base for the therapeutic use in MDD yet. Answering these outstanding questions regarding the optimal design of a treatment course is challenging because the demographic and clinical characteristics of patients might also influence the outcomes of studies irrespective of the stimulation protocols, particularly when sample sizes are small.

The present meta-analysis has several limitations. First, the studies selected here may not be representative of the large body of literature available in this field with respect to the stimulation protocols and the primary diagnoses. In the attempt to homogenise the study parameters, we focus on one stimulation frequency (10 Hz) and one outcome assessment point (after 10 sessions) rather than the clinically-approved stimulation protocols. Furthermore, studies with mixed unipolar and bipolar MDD samples were excluded since the antidepressant outcomes may differ in patients with unipolar vs. bipolar MDD. Second, the current meta-analysis is based on the outcomes of active stimulation in all studies because some studies did not have sham control groups. Although the study design (RCT vs. open-label) had little effect on the primary outcomes (depression severity), the secondary outcomes (response and remission rates) tended to be inflated in the open-label studies with either coil. Such inflated effects might have resulted from the potentially high risk of selection, performance, and detection bias in the open-label studies because no other risks were identified in all studies using the Cochrane tool [59] (Additional file 1: Table S1). Therefore, the antidepressant outcomes reported here cannot be used to quantify the magnitude of the clinical efficacy of either coil. Third, there was too little information regarding safety of both methods reported at session 10 of treatment. Thus, we could not analyse the drop-out rates meta-analytically in the current analysis. Fourth, it was not possible to match the rTMS and the DTMS studies based on the number of pulses as well as other stimulation parameters, such as intensity and inter-train-interval. It is likely that the antidepressant outcomes of either coil depend on a complex interaction between various stimulation parameters and patient characteristics [11]. Fifth, the antidepressant outcomes are based on various versions of HDRS and consequently on different definitions (cut-off scores) of the secondary outcomes (response and remission rates). Relative to the primary outcome, the analyses regarding the secondary outcomes are also based on a lower volume of data with unequal number of open-label studies and RCTs for each coil. These factors might have masked any real differences in the secondary outcomes between both coils. Sixth, our comparison of antidepressant outcomes is based on the between-study rather than the within-study effects. Research conducted independently of the companies producing various non-invasive brain stimulation devices is required to compare the efficacy of such coils in head-to-head designs. Finally, we have not assessed the antidepressant outcomes of FDA-approved clinical protocols of rTMS or DTMS and instead focused on the same frequency of stimulation (18–20 Hz). The focus on stimulation frequency contributed to exclusion of majority of rTMS studies since rTMS with F8-coil is most often conducted using the frequency of 10 Hz rather than 20 Hz [3] while 18–20 Hz is used in all DTMS protocols in MDD studies to date. The overwhelming volume of evidence suggests that rTMS or DTMS with higher frequencies produces greater beneficial effects by alleviating symptom severity and improving cognition in MDD [34, 60, 61]. However, high frequencies are also associated with the risk of seizures, scalp discomfort, and headaches [62]. In general, a definition of ‘high-frequency’ stimulation is highly variable and includes any frequency above 1 Hz [34]. As a consequence, the high-frequency protocols in rTMS studies have also utilised a wide range of frequencies (e.g. 10, 15, and 20 Hz), while DTMS has been more restrictive in terms of frequency range (18–20 Hz). Although 10 Hz stimulation is being most often used in the clinic when applying rTMS with the F8-coil, it is unclear which of these high frequencies is most effective when controlling for other potentially confounding variables. Our results suggest that each coil may require an individual combination of parameters for the most optimal antidepressant outcomes for individual patients.

When matched on the frequency of stimulation, rTMS with F8-coil and DTMS with H1-coil produced large, acute reduction in depression severity at session 10 relative to baseline in unipolar MDD. The reduction in depression severity was greater in studies with younger patients. The comparison between coils showed a larger reduction in depression severity in H1-coil vs. F8-coil studies (independent of the study design or the concurrent pharmacotherapy) and a trend towards higher remission rates in F8-coil vs. H1-coils studies. These effects are based on a low volume of studies, are not controlled for placebo, and may not be clinically-relevant. The stimulation protocols differed systematically because stimulation was more focal but less intense (80–110% of the resting motor threshold, MT) in the F8-coil studies and less focal but more intense (120% MT) in the H1-coil studies. Head-to-head trials are required to compare the antidepressant outcomes of rTMS with F8-coil and DTMS with H1-coil to identify the most optimal stimulation protocols for acute and longer-lasting efficacy.