Background
Depression is a global illness affecting 3.8-5.0% of the adult population and accounting for 280 million cases yearly worldwide [
1]. Major Depressive Disorder (MDD) is the fourth leading cause of global disease burden, especially Treatment-Resistant Depression (TRD) have significant socio-economic consequences detectable in reduced work productivity and greater health-care resource use (HCRU) [
2].
There are several effective psychopharmacological and psychotherapeutic approaches to treat MDD. However, approximately 50–60% of patients with MDD do not have adequate response to treatment or fail to achieve remission [
3,
4]. Within this heterogeneous population, the treatment of patients who meet the well-defined criteria of TRD is a major challenge. The European Medicines Agency (EMA) defined TRD as an unsatisfactory response to two adequate trials of two different classes of antidepressants at the adequate dosage for a sufficient duration [
5,
6]. The proportion of TRD among MDD is between 4 and 20% based on prescription registers and literature data [
7,
8].
Beyond currently available pharmacological and psychotherapeutic modalities, several neuro-modulation methods, including repetitive Transcranial Magnetic Stimulation (rTMS) have been applied for the treatment of TRD. rTMS utilizes an electromagnet to generate local magnetic field pulses to modulate the activity of local neural circuits and brain function [
9]. The advantage of rTMS over other therapies is - in addition to its efficacy in depression - that it may alleviate some elements of cognitive impairment in depression and is more acceptable to patients, considering its adverse-effect profile [
10‐
12].
Numerous studies and meta-analyses have demonstrated the efficacy and cost-effectiveness of rTMS in the treatment of TRD [
9,
13] but the validity and the generalizability of these results are limited by several methodological issues. These include the heterogeneous patient population, the limited sample size, the different definitions for TRD, the lack of precise definition of response and remission, the differences in the methods used to evaluate improvement, and the different rTMS treatment protocols and regimens (monotherapy or adjunctive).
To address these methodological issues, a meta-analysis was conducted to collect and analyze all available evidence regarding the application of adjunctive rTMS treatment in the management of TRD, and how different methodological issues may impact the efficacy of add-on rTMS treatment.
Methods
Search strategy and study selection
We conducted a systematic literature search in PubMed and Scopus databases for articles published between January, 2000 and March, 2022 with terms and subject headings: repetitive transcranial magnetic stimulation or rTMS, major depressive disorder or treatment resistant depression, sham or sham control, based on the following search query: TITLE-ABS-KEY (rtms OR “repetitive transcranial magnetic stimulation”) AND TITLE-ABS-KEY (“major depressive disorder” OR “treatment resistant depression”) AND TITLE-ABS-KEY (sham OR “sham control”).
We identified Randomized Clinical Trials (RCTs) that were published in English and investigated the relative efficacy of rTMS technology compared with standard pharmacotherapy in adult human population with unipolar TRD. Studies with high-frequency and unilateral application were included for further investigation. We included studies with the number of sessions ranging between 10 and 30 and the number of pulses between 6000 and 120,000. The motor threshold (MT) ranged from 80 to 120%.
Further eligibility criteria included (i) the study was randomized, sham-controlled; (ii) there was information regarding the number of therapeutic failures (and there were at least 2 failed antidepressant treatment); (iii) the severity of depression was assessed with Hamilton Depression Rating Scale (HAM-D/HDRS) [
14] or Montgomery-Asberg Depression Rating Scale (MADRS) [
15]; (iv) there was information whether the rTMS was used in monotherapy or as adjunctive treatment; (v) the localization of the stimulation was the left dorso-lateral pre-frontal cortex (DLPFC); and (vi) the range of stimulation frequency was reported in the study (5, 10, 20 Hz).
We excluded studies that evaluated different application or indication of the rTMS or if the TRD was accompanied with specific conditions or events (e.g., post-stroke depression, depression related to substance use disorder, etc.). Also, we did not include studies with no reported outcome data.
Data abstraction, validity assessment
The following information and parameters were extracted by two authors independently: publication year, study design (sham-controlled, monotherapy or augmentation), stimulation frequency, motor threshold, number of therapeutic failures, number of sessions, number of pulses, applied depression scale, age, sex, number of subjects, remission and response rates in active rTMS arm and sham rTMS arm. Any disagreement or uncertainty regarding inclusion of the study and data abstraction was resolved by consensus and feedback from a third team member.
Response rate was defined as the number of subjects that had a minimum 50% or more reduction in post-treatment scores on the Hamilton Depression Rating Scale (HAM-D/HDRS) or on the Montgomery–Asberg Depression Rating Scale (MADRS) [
14,
15].
Remission rate was defined as the number of subjects assessed as being in remission based on psychometric testing results. These varied based on the assessment tools used, however defined in terms of a cut-off point, generally as MADRS < 10 and/or HAM-D/HDRS ≤ 7 points. In contrast to MADRS, where the cut-off score for remission is less then 10 points, there was a heterogeneity between the different studies in terms of defining remission by HAM-D/HDRS. Remission was assessed by 9 studies, in which the cut-off value for remission was defined as ≤ 7 in 5 studies [
16‐
20], ≤ 8 in 2 studies [
21,
22] and ≤ 10 in one study [
23] using HAM-D/HDRS and < 10 in one study applying MADRS [
24].
Methodological quality of the included studies was graded using the Scottish Intercollegiate Guidelines Network (SIGN) checklist for controlled trials [
25]. The instrument is composed of 13 questions, which are grouped into the following sections: internal validity and overall assessment of the study. Studies are considered to present high quality when most or all the criteria are met, acceptable quality if more than half of the criteria are met, low quality if fewer than half of the criteria are met and unacceptable if the paper is not relevant to key question. The quality assessment was performed by a reviewer under the supervision of a senior professional, and discrepancies were discussed until an agreement was reached.
Statistical analysis
We performed the statistical analysis in RStudio Version 2022.02.0. The proportion of patients in the response and the remission groups were converted to odds ratios (ORs) and risk ratios (RRs) with 95% confidence intervals (CIs), where values beyond 1 reflect greater improvement for patients in the treatment group. For reporting the results, we have selected RR over OR due to its enhanced informativeness considering the study design and its greater convenience for interpretation in the context of health-economic studies [
26]. We assessed data heterogeneity using Cochran Q statistic and we measured the proportion of variance accountable to data heterogeneity using I
2 statistic. Data were synthesised using random-effects model, which accounts for possible heterogeneity of the data. The year of publication, number of sessions, number of pulses, and the interaction of number of pulses and sessions were included as moderator effects in the models to test for the effects of study characteristics on outcomes. Effect sizes and weights were visualised in forest plots. Publication bias was assessed with Begg and Mazumdar’s rank correlation test and visualised using funnel plots, in which effect size is plotted against standard errors estimated from study size [
27].
Discussion
The results of the presented meta-analysis of 19 studies revealed that nearly 40% of TRD patients responded to the adjunctive active rTMS treatment, while with sham-rTMS (placebo group) this rate was just over 10%, meaning that the response rate was almost two and a half times more with active rTMS treatment (RR: 2.25, NNT: 6.1). This difference was even more pronounced when analyzing remission rates. Based on the 9 analyzed studies, nearly 36% of the TRD patients with rTMS treatment achieved remission, while barely 8% in the placebo group (sham-rTMS). Those with active rTMS treatment were almost three times more likely to achieve complete remission than those who did not receive active rTMS treatment (RR: 2.78, NNT: 5.4). An even greater difference was found when assessing the latter 9 studies for rTMS responders and remitters with RR of 1.16 and 4.27, respectively (Table
2.). These ratios show similarities with the results of previous meta-analyses however, the effect size is slightly smaller than in the former published literature [
9]. Nevertheless, the value of these presented results is underlined by the fact that the strict eligibility criteria and the sophisticated statistical methods have helped to overcome the major methodological issues (e.g., heterogeneous patient population, differences in rTMS protocols and methods used to evaluate improvement) that limited the validity and generalizability of the conclusions from previous analyses [
9]. The potential contributing factors in the background of the differences in the effect sizes in the current and the previous analyses are discussed below.
Previous meta-analyses had some limitations we aimed to eliminate in this study. Major limitations are associated with the
heterogeneity of the study populations and the rTMS protocols applied across different studies. Although in 2008 FDA approved rTMS monotherapy in MDD [
38], recent clinical practice applies rTMS generally as an add-on treatment to antidepressants. In this analysis, only those studies were included in which rTMS was used as an adjunctive treatment to antidepressants. Furthermore, in the past years, therapeutic guidelines included rTMS as a treatment alternative in patients with MDD, who did not respond well to antidepressant medication(s) [
39]. Therefore, we investigated the impact of rTMS treatment only in patients with TRD, defined as an inadequate response to two previous antidepressant medications. Applying these inclusion criteria above provided a more homogeneous patient population and reflects everyday clinical practice.
Other major limitation of previous studies is the
heterogeneity of rTMS protocols [
40]. In this meta-analysis, we focused on the unilateral, high-frequency repetitive TMS technique, and included studies only with predefined specific rTMS parameters (such as 5–20 Hz, 10–30 sessions, 6000-120000 pulses, and 80–120% MT). However, similarly to Leggett et al. (2015) our deeper analysis revealed that there are no statistically significant differences between the various modalities of the rTMS technology, and the differences in the settings did not significantly affect the RR related to the efficacy [
41].
The interpretation of several clinical trials using rTMS is also
limited by their sample size and composition [
40]. Furthermore, there is usually missing or incomplete information regarding the characteristics of subjects and their depression. Bipolar depression or mixed features in depression can be a confounding element as it can be an important reason for treatment resistance, and these patients may need more time to achieve remission [
42,
43]. Hypothetically, subgroup analysis for bipolar depression may be a solution for this issue, however, with the currently available studies, this may lead to sample sizes unsuitable for meta-analysis. Therefore, Fornaro et al. (2020) recommended to set a proportion of patients with TRD and/or bipolar disorder as threshold for inclusion in future clinical trials [
43]. This may also apply for the elderly population, however different rTMS protocols may be used for this population [
44,
45].
Sham methods and blinding techniques in clinical trials with rTMS are other potential sources of bias as the different coil settings in degree, distance and rotation may influence the magnitude of placebo effect, generally resulting in smaller effect size difference between the active and sham group [
46]. This, and the evolution of the blinding techniques can be an origin for the smaller effect size found in our study. The inclusion of sham coil (identical to the active coil), or configurations that could control both active and sham devices are generally common nowadays [
24] while more than 10 years ago some studies included rTMS treatment naïve subjects, as control group [
36]. Furthermore, in many studies the information regarding the blinding techniques is lacking or incomplete, stating that the evaluations were carried out by blinded raters [
16‐
24,
28‐
33,
36,
37], or only mentioning double-blind study design [
34,
35].
The primary aim of this meta-analysis was to assess the relative efficacy of rTMS treatment in TRD. Another purpose of conducting the meta-analysis was to use its results in analyzing the cost-effectiveness of rTMS compared with standard care as part of a health technology assessment (HTA) process to support decision-making on whether to include rTMS technology in public reimbursement. Therefore, the potential risks of publication bias were minimized with stringent inclusion criteria and with the application of various validated statistical methods to analyze heterogeneity and publication bias. The lack of significant heterogeneity and low publication bias makes our results reliable and may be used in cost-effectiveness analyses (CEAs) [
47].
However, in contrast to a health technology assessment published in 2021 [
48] our results showed greater risk ratio (efficacy) in remission than in response. The possible reason behind this different finding may be due to the more homogenous patient population included in this meta-analysis, as it may take longer for patients with bipolar disorder to reach complete remission compared to patients with MDD. Furthermore, our results may implicate a similar tendency as in the literature [
49], that
patients who are not responding initially to rTMS may not reach remission, however others with good initial response are more likely to achieve complete remission. This means, that early response to rTMS treatment may predict full remission with full-length (20–30 sessions) treatment. According to this, the assessment and the evaluation of the initial response to rTMS treatment may be included in the rTMS protocols to find the patient population who will most likely benefit from the rTMS treatment.
These results also implicate, that there is a subgroup of patients, who responded well to rTMS adjunctive treatment, meaning that they are not only responding to the treatment with add-on rTMS, but they may reach complete remission. From a clinical perspective, it may indicate that this sub-group of patients with partial response to antidepressants can achieve complete remission with adding rTMS. With other words, ‘rTMS may push these antidepressant partial responders to complete remission’.
It is interesting to interpret our results in the light of a recent meta-analysis of
antidepressant treatment in terms of efficacy compared to placebo, published by Cipriani et al. in 2018. Based on 522 trials, the OR ranged between 1.15 and 1.55 for different antidepressants against placebo [
50]. The relative efficacy of rTMS in this patient group is comparable or even better than antidepressant therapy. That is important, because patients with TRD are less likely to respond to further antidepressant therapy, whether it belongs to another pharmacological class or not. For these patients, neuromodulation technologies, like rTMS may be a rational treatment option as an add-on therapy or even in monotherapy [
51].
The main limitations of our findings are originated from the aforementioned weaknesses of the RCTs on rTMS treatment and the controversies regarding the definition of TRD. The latter was based on the number of unsuccessful antidepressant therapies, as defined by the EMA. The blinding technique during the sham rTMS was not incorporated in our analysis, as most of the clinical studies did not provide detailed information or description on it. The inclusion of different studies with a slightly varying cut-off points and duration of remission may contribute to the small difference between remission and response rates, and it should be also highlighted as a limitation. Further shortcoming is that several former studies (especially before the year of 2010) did not separate between patients with bipolar and unipolar depressive disorders and the ratio of patients with bipolar depression in the study population varied from 8 to 50% [
16,
22,
30,
31,
35,
36], which may result in smaller effect of the rTMS treatment. In this analysis, studies included more than 20% patients with bipolar disorder were excluded.
The results of this meta-analysis based on recent RCTs proved that rTMS treatment may be used effectively in the treatment of TRD, as it offers the possibility of achieving not only a symptomatic response, but also a complete remission in patients, who respond well to the treatment. In order to obtain reliable information regarding the effective applicability of rTMS in different subgroups of patients with MDD and to identify the patient population that would benefit most from rTMS treatment, further research is needed in this field with more precise methodology, including standardized treatment protocol and multi-centre study design, with further subgroup analyses in larger numbers of subjects and with possible predictor assessments (e.g. biomarkers, baseline cognitive performance) [
52,
53]. Furthermore, recent studies and consequently the meta-analyses include treatment response and remission as primary outcome measures, however functioning and quality of life (QoL) would be also suitable measures to assess the efficacy and also the effectiveness of rTMS [
40].
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.