Repetitive transcranial magnetic stimulation for the treatment of suicidal ideation in a naturalistic setting

Dear Editor,

Suicidal ideation (SI) is a major global health concern, with a lifetime prevalence of 9 % [1,2]. SI is associated with severe psychopathology and morbidity [3–5]. Repetitive transcranial magnetic stimulation (rTMS) shows promise as a treatment for SI in patients with treatment-resistant depression (TRD), yet there is limited evidence guiding the use of rTMS in this population [6]. In this letter, we describe a retrospective analysis of our clinical database that provides evidence for the superior efficacy of theta burst stimulation (TBS) compared to 18 Hz deep TMS for the treatment of SI in patients with TRD. We also describe predictors of SI response and remission, as well as trajectory of improvement.

We retrospectively reviewed the medical records (Epic, Verona, WI, USA) of 132 patients with TRD and SI that underwent rTMS at the UC San Diego Interventional Psychiatry Clinic between May 2017 and May 2024. Most patients received either unilateral or bilateral theta burst stimulation (TBS; N = 67; Mean Sessions: 38.9), 18 Hz deep TMS (18 Hz dTMS; N = 53; Mean Sessions: 35.5), or TBS with a deep TMS coil (TBS dTMS; n = 10; Mean Sessions: 38.4). rTMS protocols were generally chosen based on provider preference and availability. Details regarding the rTMS protocols and inclusion/exclusion criteria can be found in the supplemental methods.

The SI item on either the Patient Health Questionnaire 9 (PHQ-9) or Beck Depression Inventory (BDI-II) [7] was used to quantify SI. Both item scores range from 0 to 3 and were aggregated as: 0 = none, 1 = mild, 2 = moderate, 3 = severe. Patients without SI at the end of treatment (score = 0) were classified as “SI-remitters”, and those with an improvement ≥1 were classified as “SI-responders”.

Sociodemographic and clinical characteristics of the sample can be found in the Supplemental Table 1. Patients tended to be middle-aged (M = 55.9, SD = 17.57) Caucasian (67.4 %) females (58.3 %), with severe depression (PHQ-9 M = 20.1 ( ± 4.8)). 35.6 % (n = 47) reported a past suicide attempt, and 86 % reported at least one comorbid psychiatric diagnosis.

According to Wilcoxon signed-rank test, symptoms of SI significantly decreased from baseline (M = 1.63, SD = 0.81) to post-rTMS treatment (M = 0.79, SD = 0.97; Z = 8.098, p < 0.001; Cohen’s d = 0.70, 95 % CI [0.53, 0.88]). Following TMS, 48 % of patients achieved SI-remission and 61 % SI-response. SI-remission rates among patients with mild (55 %) and moderate SI (48 %) at baseline were similar, while remission rates for patients with severe SI (score = 3) at baseline was 32 % (Supplemental Fig. 1a). However, baseline SI was not a significant predictor for SI remission according to logistic regression (X²(2) = 4.20, p = 0.122). 69 % of patients showed convergence between depression response (≥50 % reduction in PHQ-9 total score) and resolution of SI symptoms, experiencing either both or neither. However, 11 % of patients experienced response of depression without SI remission, and 20 % experienced SI remission without depression response.

55 % of patients who received TBS (pre: M = 1.90, SD = 0.91; post: M = 0.85, SD = 1.13) and 40 % of patients who received 18 Hz dTMS (pre: M = 1.28, SD = 0.57; post: M = 0.75, SD = 0.78) achieved SI-remission (Supplemental Fig. 1b). A Linear mixed effects model (LMM) was implemented to test for differential effects of rTMS protocols on SI remission (details in Supplemental Methods). Changes in SI symptoms differed across treatment protocols (F_(2,127) = 6.01, p = 0.003) (Supplemental Fig. 1b), with greater improvements in the TBS and TBS-dTMS groups compared to the 18 Hz dTMS group (TBS vs. 18 Hz dTMS: t = 3.07, b = 0.516, p = 0.003, 95 % CI [0.184, 0.849]; TBS-dTMS vs. 18 Hz dTMS: t = 2.45, b = 0.772, p = 0.016, 95 % CI [0.148, 1.395]).

Overall, there was a near-linear decrease in SI symptoms for the first 22 sessions, with a subsequent plateau (Fig. 1a). A second LMM was implemented to detect when treatment effects diverged between SI-responders and non-responders (details in Supplemental Methods). The trajectory of SI symptoms throughout the treatment course differed significantly between SI-remitters and non-remitters (F_{(8, 867)} = 12.4, p < 0.001) (Fig. 1a). SI symptom severity of patients achieving SI-remission and those without remission differed between session 2 to 6 (Z = −2.88, p = .004), 7 to 11 (z = −4.79, p < .001), 12 to 16 (z = −5.08, p < .001), 17 to 21 (z = −5.72, p < .001), 22 to 26 (z = −5.97, p < .001), 27 to 31 (z = −7.41, p < .001), 32 to 36 (z = −7.50, p < .001), and sessions 37 to 41 (z = −8.49, p < .001) (Supplemental Fig. 1c). At baseline, SI symptom severity did not significantly differ between SI remitters and non-remitters (z = −1.88, p = .061). All significant results remained significant with Bonferroni correction (0.05/9 = 0.0056). Additionally, for patients not experiencing SI-response at a given time-point during treatment, the proportion of these patients who would experience remission decreased drastically to 17 % by treatments 17–21 (Fig. 1b). This could inform treatment selection for patients experiencing ongoing suicide risk late into their rTMS course.

Fig. 1.

a) Trajectory of SI symptoms across rTMS treatment for SI-remitters and non-remitters b) Remission rates for SI non-responders at a given number of rTMS sessions.

Penalized logistic regression was used to identify pre-treatment predictors of treatment response (see details in Supplemental Methods). A model with 30 pre-treatment variables did not predict SI-response (AUC = 0.50) or SI-remission (AUC = 0.57) better than chance. Therefore, no robust predictors of treatment outcome were identified. Future work should explore neurobiological biomarkers of treatment response.

This study has several limitations. The rTMS interventions described are FDA-approved for depression and not primarily intended to target SI. This study consists of retrospective data from a naturalistic clinical setting, introducing potential biases from placebo effects, selection bias, and other confounding variables. SI severity was quantified using single items from the PHQ-9 or BDI scale, offering a limited view into the spectrum of SI. Additionally, the PHQ-9 assesses the frequency of symptoms, while the BDI focuses on severity. However, both items range from 0 to 3, and concordance between the two SI measures was high among 11 patients repeatedly assessed with both scales (r = 0.82).

This is the first study to suggest the superior anti-suicidal effects of TBS compared to 18 Hz dTMS. This is in line with prior studies demonstrating the effectiveness of TBS for the treatment of SI [6,8,9]. Our results suggest that the proportion of patients achieving remission of SI is limited beyond sessions 17–21 in a standard rTMS course. In summary, this retrospective cohort analysis offers novel insights into the use of rTMS as a treatment for SI in TRD, with a focus on the effects of various rTMS protocols, symptom trajectories, and predictors of treatment outcome. Future clinical trials should explore further optimization of TBS paradigms for the treatment of SI.

Supplementary Material

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.brs.2025.01.005.