Introduction

Major depressive episodes (MDE) represent a major public health problem associated with increased functional disability and mortality. Limited efficacy of antidepressants [1] have triggered a specific interest in new therapeutics such as noninvasive neuromodulation, including repetitive transcranial magnetic stimulation (rTMS).

rTMS is the most widely used and studied form of TMS and uses an alternating current passed through a metal coil placed against the scalp to generate rapidly alternating magnetic fields, which pass through the skull nearly unimpeded and induce electric currents that depolarize neurons in a focal area of the surface cortex [2–4]. Although the mechanism of action is still debated, one hypothesis is that stimulation of discrete cortical regions alters pathologic activity within a network of grey matter brain regions that are involved in depression and connected to the targeted cortical sites. This is supported by functional imaging studies that show rTMS can change activity in brain regions distant from the site of stimulation [5]. Furthermore, rTMS has many molecular effects, including increased monoamine turnover and normalization of the hypothalamic pituitary axis [6]. In one neuroimaging study of depressed patients, a prefrontal serotonin deficiency at baseline normalized after treatment with rTMS [7].

The effect of rTMS appears to vary according to the frequency; high frequency (HF) stimulation is thought to excite the targeted neurons (and is typically used to activate the left prefrontal cortex), whereas low frequency (LF) stimulation appears to inhibit cortical activity (and is usually directed at the right prefrontal cortex) [8]. Consistent with this hypothesis, a review examined 66 studies in depressed patients who were treated with TMS targeting the dorsal lateral prefrontal cortex and found that HF TMS generally increased regional cerebral blood flow and that LF TMS generally decreased regional cerebral blood flow [9].

Numerous meta-analyses of randomized trials have found evidence that rTMS provides a clinically relevant benefit to patients with treatment resistant depression (TRD) [4,10–12] such as Brunoni et al. who identified in a 2017 network meta-analysis that the rTMS was more effective than sham stimulation. However, most studies in the rTMS literature are underpowered, with a risk of excess significance bias [13,14]. As a consequence, the use of rTMS for MDD in clinical practice is still not officially recommended in several countries such as France. Conversely, in the United States of America, repetitive TMS has been approved by the Food and Drug Administration (FDA) to be used in a routine clinical setting for depression. In the United Kingdom, the National Institute for Health and Care Excellence (NICE) adopted a more nuanced position regarding its efficacy arguing that “the evidence on its efficacy in the short-term is adequate, although the clinical response is variable” [15]. Such divergent opinions about the evidence-based efficacy of this technique may in part reflect the numerous challenges in developing and validating such a complex stimulation technique which involves numerous parameters that can impact its efficacy and safety [16–18].

The accurate neuroanatomical location of the CPFDL is surely one of these parameters. While it is now considered as consensual across studies [19–21], it is challenging to specifically reach this region due to the interpersonal variability in terms of size and shape. Several methods have been developed recently to enable better targeting, ranging from the simplest methods (Standard method, 10–20 EEG localization) to increasingly sophisticated ones (neuronavigation with structural MRI, neuronavigation with functional MRI, Electric field modeling). Despite advances in targeting techniques, the superiority of one method over another remains, to date, unestablished, especially in routine clinical practice [19,20,22,23].

The standard method (1995) stipulates that for each patient the CPFDL region is over the left prefrontal cortex, determined by moving the TMS coil 5 cm anterior to the motor threshold location along a left superior oblique plane with a rotation point about the tip of the patient’s nose [24]. Using a neuro-navigation system to perform a comprehensive quantification of target localization errors, Nauczyciel et al. [25] have identified 3 sources of error with this method: (i) the cap repositioning, (ii) the interexpert variability in coil positioning and (iii) the distance between the stimulated point and the expected target. Moreover a recent study [26] compare standard and neuronavigation targeting methodologies on distance, angle, and electric field magnitude values demonstrated that neuronavigation technology enables more precise and accurate TMS positioning, resulting in the intended stimulation intensities at the targeted cortical level. Despite this difference, the authors highlight the need for prospective studies to determine the clinical superiority of one method over the other.

To our knowledge, only one study has compared the two approaches in a randomized controlled manner. Fitzgerald et al. designed a randomized trial of repetitive TMS targeted with magnetic resonance imaging (MRI) based neuro-navigation in TRD. In this study including 51 patients with treatment resistant depression the neuro-navigational approach targeting a specific site at the junction of Brodmann areas 46 and 9 in the DLPFC based on each individual subject’s MRI scan outperformed the standard localization group at 4 weeks [19].

We therefore designed this study to reproduce the findings from Fitzgerald et al. [19] in the French context using a larger sample size and an in-house developed French neuro-navigation system (Syneika; syneika.fr) which takes its originality in the automatic location of the left DLPFC. Thus, our definition of reproducibility includes both the concept of result reproducibility and generalizability, as outlined by Goodman et al. [27]. A first objective of this multicenter randomized controlled trial was to demonstrate the superiority in terms of efficacy over the depressive symptoms of HR-rTMS of the left DLPFC targeted with MRI based neuro-navigation versus the standard localization technique in MDD using the MADRS score.

Methods

Results

Participant flow

105 participants were recruited starting in January 2013 (21/01/2013). Last visit of the last patient was on April 2017 (13/04/2017).

The mITT population comprised 92 patients (88% of the randomized patients with 45 (87%) in the neuro-navigation group and 47 (89%) in the standard group). The per-protocol population comprised 74 patients (70% of the randomized patients with 32 (62%) in the neuro-navigation group and 42 (79%) in the standard group). Fig 1 summarizes the participant flow in this study. Table 1 displays initial characteristics of both groups.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Demographic and baseline clinical variables (n = 102).

https://doi.org/10.1371/journal.pone.0317597.t001

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Study flow chart.

https://doi.org/10.1371/journal.pone.0317597.g001

Efficacy results

Primary outcome.

No difference was evidenced (p = 0.58) in terms of treatment response with 14 (31.8%) responders in the intervention group, and 16 (35.6%) in the control group (relative risk, 0.89; 95% CI, 0.50 to 1.61 and absolute risk difference, 3.7%; 95% CI, -0.16 to 0.23]. Similar results were observed in the per-protocol population (Table 2).

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Response and remission criteria at D14 and D44.

https://doi.org/10.1371/journal.pone.0317597.t002

Secondary outcomes.

No difference was evidenced for response at both D14 and D44 as displayed in Table 2. Difference in means was of 0,01; 95% CI, -3.06 to 3.26 for BDI at D44 and of 0.11; 95% CI, -2.42 to 5.02 for ERD at D44. Fig 2 displays results observed on MADRS (Fig 2A), BDI (Fig 2B) at each time point.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Mean scores (+/- Standard deviations) at baseline and at each visit (D1, D14, D44) in (a) Montgomery and Åsberg Depression Rating Scale, (b) Beck Depression Inventory and (c) Retardation Rating Scale for Depression scale.

https://doi.org/10.1371/journal.pone.0317597.g002

Safety results.

Rates of adverse events were similar in both groups with 23 (44%) and 1 (2%) of adverse events and serious adverse events in the neuronavigation group versus 20 (38%) and 0 (0%) in the standard group (p-values of 0.55 and 0.49 respectively). Table 3 presents all adverse events.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Safety results.

https://doi.org/10.1371/journal.pone.0317597.t003

Discussion

Our study failed to reproduce Fitzgerald et al.’s study result supporting the use of neuro-navigation system to optimize rTMS efficacy [19]. Several reasons could explain our inability to reproduce such result. First, Fitzgerald et al.’s study was performed on a single clinical center and we do not have information on the number of clinicians that administered rTMS treatment and their expertise. Our study is subject to greater variability in rTMS operators due to its multicentric design, involving multiple clinicians at each center and reflecting real clinical conditions with differences in equipment and timing. In their 2022 study, Caulfield et al. [26] compared neuronavigation with the standard placement method, assessing the influence of TMS operators on elastic cap placement during each stimulation day. They examined its effects on the distance, roll/pitch angle, and yaw angle deviations from the DLPFC target. The results revealed significant differences in cap placement measurements between operators using the standard method, but no such differences were observed with neuronavigation. Neuronavigation demonstrated consistent accuracy, unaffected by operator variability. Overall, these findings emphasize the operator’s influence on the standard method, which may contribute to variability in our study.

Second, Fitzgerald’s study included a relatively small sample size, with 24 patients in the neuro-navigational group and 27 in the standard group. While our study involved a larger sample, it remains relatively small, as it was designed to detect a somewhat large effect of neuro-navigated rTMS. A lack of statistical power may therefore explain our negative results. In addition, the modified ITT analyses primarily reflect the implementation of the neuro-navigation system in our study and the constraints encountered during its execution. Notably, we faced several deviations, including significant delays between randomization and Day 1 due to technical and logistical challenges. These delays may have led to spontaneous improvement in some patients before the first session, potentially resulting in a less severe patient population compared to other studies. Despite these factors, the per-protocol analyses did not demonstrate the superiority of neuro-navigated rTMS. Importantly, the sample size in this study population remained larger than that of Fitzgerald’s trial, which reported a very large effect size with 10/24 patients (42%) in the targeted group and 5/27 (18%) in the standard group met the response criteria at 4 weeks.

Third, the stimulation parameters were noticeably different in both studies. The rTMS treatment was applied at 10Hz with an intensity of 100% of resting motor threshold for a total of 20 sessions per patient in Fitzgerald et al. In comparison, in our study the rTMS treatment was applied at 20Hz with an intensity of 110% of resting motor threshold for a total of 10 sessions per patient. In particular, the limited number of sessions conducted in our study may have influenced the operator-dependent effects in standard positioning and it perhaps prevented a more pronounced differentiation between groups, as observed in the study by Fitzgerald et al. Indeed, previous research [33,34] supports the idea that a greater number of sessions can enhance overall efficacy.

Forth, the differences in neuro-navigation system could also explain why we found no difference between the two groups in our study. The localization method of the DLPFC used in Fitzgerald study is based on warping scan into Talairach space (by using SPM99), and then reporting target spot already defined in this space (-45;45;35) into the warped scan. In our study, the Syneika device’s localization method implies a template brain scan (named Colin27) on which the targeted spot (DLPFC) has been localized by a neuroanatomy expert. Patient’s brain, automatically segmented from MRI scan, is then registered to Colin27 template thanks to rigid and non-rigid transforms. These transforms are finally used to report the targeted spot identified on Colin27 template scan to the patient’s MRI space. Using our method, we have estimated the error of this method at 12mm root mean square compared to the same localization manually done on each patient’s MRI by a neuroanatomy expert, while no error estimation was calculated in Fitzgerald study. Furthermore, in Fitzgerald study, the patient’s head is placed in a frame to prevent movements during the treatment, while in our study, no such movement preventing solution was used; the patient was only seated in a chair on which he was able to release his head. Our device was nevertheless able to detect large and lasting movements.

A final point that may explain the difference lies in the clinical characteristics of the patients. In our study, we included patients with more heterogeneous types of depression than in Fitzgerald’s study. Twenty percent of our patients had bipolar depression, which was an exclusion criterion in the Fitzgerald study. A recent meta-analysis [33] suggests that high-frequency stimulation of the left DLPFC may not be optimal for bipolar disorder, potentially impacting overall efficacy. Despite these potential differences, we chose not to conduct subgroup analyses, as they carry a risk of alpha risk inflation and cannot be reliably interpreted when not planned a priori and/or when the results are negative, as in our case. The type of depression may have a different cerebral impact, affecting brain structure, function, and connectivity, which may influence target selection and response to treatments like rTMS [11,35]. It is important to note that the 5 cm rule does not always target the classic DLPFC areas (BA 9 and 46) [36]; instead, it reaches a transitional region of the DLPFC connected to the Default mode network and motor cortex. This could explain the different, and sometimes beneficial, responses to rTMS in a heterogeneous clinical population.

It is important to note that our result is, however, consistent with other studies that compared neuronavigation versus the other method. Hebel et al., 2021 [22] conducted a single-center, two-arm, randomized, double blind study with 37 inpatients assigned to receive either neuronavigated iTBS or 10–20 EEG system-based F3 guided iTBS over the left dorsolateral prefrontal cortex (lDLPFC). The study did not demonstrate clinical superiority of neuronavigated localization compared to the 10–20 EEG system-based F3 approach. In another hand Tsukuda et al., 2024 [23] examines the effectiveness of three targeting methods for rTMS in treating depression among Japanese patients: the standard 5 cm rule, the F3 method, and neuronavigation. The study compares these approaches to determine their clinical efficacy and accuracy in targeting the lDLPFC in 16 depressive patients. Their results indicate that neuronavigation offers higher targeting precision, potentially leading to more consistent treatment outcomes, though differences in clinical effectiveness between methods were limited.

In conclusion, this study failed to identify any difference of efficacy of rTMS for depression when comparing neuronavigation-based targeting with the standard method. While neuronavigation theoretically offers more precise targeting of the DLPFC, no operator dependence, less target deviation [26], and a better distribution of the eclectic fields [37] this did not translate into superior clinical outcomes in our sample. Targeting methods have evolved considerably over time, enabling a personalized approach to depression treatment, particularly through functional MRI neuronavigation and electric field modeling. Despite the appeal and high precision of these approaches, their use in routine clinical practice remains limited due to cost and accessibility. As a result, establishing the most efficient localization method remains a clinical challenge.

The contradictory results between Fitzgerald et al. [20] and our study, as well as the limitation of both trials, should encourage the design of larger studies with optimal stimulation parameters and diverse patient populations. It would help clarify if specific subgroups could benefit more from one targeting method over the other.

Supporting information