The BRIGhTMIND study is the UK’s largest ever randomised controlled trial of TMS for TRD and the first adequately powered trial of TBS with outcomes at 26 weeks.
Clinical efficacy
Contrary to the hypotheses, there were no statistically significant differences between cgiTBS and rTMS on the primary and secondary clinical outcomes across 26 weeks, demonstrating that cgiTBS does not show superior clinical efficacy compared to MRI-neuronavigated rTMS. The primary outcome analysis of the HDRS-17 demonstrated clinically substantial improvements for both treatment arms at the final 26-week follow-up.141 There were substantial improvements in self-rated depression measures and clinically important improvements in anxiety, cognition, function, and quality of life over 26 weeks. The results also indicated that both rTMS and cgiTBS are efficacious in terms of achieving a response (50% drop in HDRS-17 score) in 30–35% participants with TRD at 26 weeks; about 65–70% obtained much less benefit or no benefit. About a fifth to a quarter of participants with TRD might achieve a remission after treatment that is still sustained for 6 months. Severity of depression, severity of anxiety, and treatment resistance were predictors of depression score after TMS, but none of them distinguished whether participants would respond to cgiTBS or rTMS. Age, gender, comorbidity, or previous experience of childhood trauma neither predicted depression score after TMS or treatment group. From a clinical perspective, a group of otherwise unresponsive patients with TRD could obtain a sustained remission, with a proportion of patients potentially remaining well with 1–2 courses of rTMS or iTBS per year.
Two trials that are in closest comparison to ours are the THREE-D23 and THETA-DEP,28 both of which compared iTBS versus rTMS, with both their treatment arms using structural MRI neuronavigation. This study’s response and remission rates were largely consistent with those observed in the THETA-DEP study, which at the end of treatments showed response rates of about 33–37% and remission rates a little lower than the present study’s of 15–19%, with depression symptom improvement maintained over 6 months.28 Thus, while prior evidence suggests that the beneficial effects of rTMS on mood in TRD may be relatively short, lasting only 1–3 months, both TMS protocols actually appear to maintain improvement for over 6 months post treatment.9,17 Additionally, the THREE-D study reported rates of 40–50% for response and 20–30% for remission, which are slightly higher rates than what we found.23 However that trial provided up to 30 TMS sessions for a number of their participants, rather than the set 20 sessions in the present study.
The patient global impression of change showed that the proportion of participants experiencing a benefit continued to increase even at the 19th and 20th sessions, and the qualitative work also indicated variability in symptom improvement. These findings taken together with the THREE-D study suggest that for a number of participants, additional sessions in those who started to make a response may have resulted in better response, which could have been observed had the present study allowed for a longer course of treatment. In support of this suggestion, a recent study indicated different response trajectories following TMS, whereby there was a rapid response group who showed dramatic improvement after 2 weeks of treatment, and two intermediate groups which showed a slower, linear improvement over 6 weeks of treatment. Importantly the two intermediate groups showed no apparent plateau of improvement.147 Therefore, from a clinical perspective the duration of a TMS treatment course for TRD might be decided on a case-by-case basis, but would normally last at least 20 sessions, possibly extended to 30 sessions if there is some response for example, 25% drop in HDRS-17 from baseline by the end of the TMS course. A further RCT to inform clinical practice for TMS in people with TRD would be a RCT offering 20 iTBS or rTMS sessions and an additional 10 sessions if there is a 25–49% drop in HDRS-17 or MADRS score at the end of treatment versus offering everyone either 20 or 30 sessions of TMS with outcomes at 16 and 26 weeks.
We provided 3000 pulses per session for both rTMS and cgiTBS, with the iTBS protocol including five runs of 600 pulses with 5-minute intervals to ensure that the duration of sessions were the same for both treatment arms for the purposes of blinding, as per the pilot work.43 An advantage to this was that the duration of therapeutic contact, the number of pulses per session, and the number of sessions were matched. Thus, non-specific effects of TMS were matched between the treatment arms. The results were largely comparable with THETA-DEP and THREE-D, both of which used 600 pulses of iTBS per session, which indicates that increasing pulses up to 3000 per session does not lead to better response rates but might be associated with higher remission and sustained response and remission rates, compared to 600 pulses per session.23,28 From a clinical perspective, the standard 600 pulse iTBS is advantageous in the respect that it requires a shorter duration of administration, meaning more patients can be treated per day, and it can do so without compromising clinical efficacy. However, we also should acknowledge that giving more pulses per session may have been superior, but the 5-minute intervals between runs might have diluted clinical efficacy.
Research suggests that dosage and time intervals between TBS protocols may affect meta-plasticity either by reducing or reversing the effect of synaptic plasticity or by increasing the effect of synaptic plasticity. Studies of healthy controls have shown that the delivery of two iTBS sessions to the motor cortex (with 5-minute intervals between sessions), led to a reduction in excitability148 or no change in excitability,149 whereas 15-minute intervals increased excitability.147 Additionally, a study of 3 runs of standard iTBS with 15-minute intervals have been found to significantly increase cortical excitability compared to 2 runs or 1 run,150 although no studies have explored clinical or adverse effects following 5 runs of iTBS until now. It is difficult to ascertain how translatable these findings are to depression and TMS delivery to the lDLPFC. However, a recent study did find that motor cortex rTMS induced modulation of corticospinal excitability predicted antidepressant response to lDLPFC 10 Hz rTMS in MDD.151 Therefore, it is plausible that the 5-minute gaps between the present study’s iTBS runs could have impeded a potential cumulative effect of cortical excitability, limiting the therapeutic efficacy of the cgiTBS to the first 600 pulses. Since the pilot intervention of cgiTBS with five runs of 600 pulses with 5-minute intervals seemed more effective than rTMS, we did not alter the iTBS intervention in the light of the emerging data.148
In addition, the moderator analyses demonstrated that higher baseline HDRS-17, higher baseline GAD-7, higher treatment resistance and completion of < 20 sessions as planned, predicted greater depression symptoms over 26-week follow-up, which is consistent with prior research.152,153 It is suggested some patients could respond to iTBS but not rTMS or vice versa,22 a view supported by members of the LEAP group. However, the findings indicate that the moderators of choice did not influence the relationship between treatment arms and depression response. It may be the case that the relationship between clinical predictors and response to different TMS protocols could be more nuanced than what can be observed at a group or population level, and instead we could explore clinical predictors based on distinct trajectories of response, as well as investigating other potential predictors such as benzodiazepine and other psychotropic medication use, number of depression episodes, depression durations, fMRI connectivity, site of stimulation, and THINC-it response. Although an association between the different rTMS and iTBS protocols and response trajectory was not found, characteristics of membership in a rapid response group included older age, lower baseline depression scores, and absence of benzodiazepine use, whereas characteristics of membership to a non-response group included higher baseline depression scores.147 Therefore, further research on the exploration of response trajectories is warranted.
The present study also found that alongside improvements in depression, there were also clinically important improvements in anxiety symptoms and quality of life, which also appear to be maintained over 6 months, in line with the findings of the THETA-DEP study.28 A further strength of the present study was identifying that there were cognitive improvements over time for sustained attention, executive functioning, working memory, and a reduction in inter-individual variability for sustained attention. The findings are consistent with a recent meta-analysis which found rTMS led to modest cognitive enhancing effects specific to psychomotor speed, visual scanning, and set-shifting ability in MDD patients.11 The absence of a placebo arm means that at least part of the improvement in cognition over time may be a practice effect. However, given that improvement in depression was a predictor of improved sustained attention, subjective cognitive performance, and a reduction in individual variability for sustained attention, this would imply that some of these improvements were a consequence of the TMS treatments, and that as depression symptoms improved so did cognitive performance.
It is also important to note that we did not match the percentage rMT (80% baseline rMT for cgiTBS and 120% baseline rMT for rTMS, as per the pilot work).43 However, 120% baseline rMT is part of the extensively used rTMS protocol,154 and TBS protocols with subthreshold rMTs may have better clinical efficacy.21 In some further preparatory work at one of the sites leading to the trial, we found that iTBS at > 80% rMT led to more participants wanting to withdraw from treatment before the course of iTBS was completed. In addition, while the study found that more participants needed to reduce motor threshold in the rTMS group, we do not believe the chosen motor thresholds would have impacted the clinical efficacy of either of the treatment groups.
As noted in Chapter 2, prior to the revision of the primary outcome analysis and sample size, the original primary outcome was to determine efficacy of cgiTBS at 16 weeks (primary clinical outcome, 50% drop in HDRS-17 score from baseline to 16 weeks). However, the responders’ results at 16 weeks show a difference of < 1% between the two treatment arms, and it would be very unlikely for this finding to be different even if we had recruited the whole of the original sample. There was good inter-rater reliability among research assessors, blinding was excellent with few unblindings, in the vast majority of cases the site of stimulation of TMS varied by < 1 cm and the angle of stimulation < 10 degrees from the target site and over the course of 20 sessions, and over 90% of the sample received 20 TMS sessions close to the site of stimulation in under 6 weeks. Therefore, the findings are unlikely to be due to the quality of the trial.
Given the lack of a sham control, we cannot rule out that some of the responses may have been due to non-specific factors such as regression to the mean, expectancy effects and hope (as demonstrated in the qualitative data), change in routine (as shown in the qualitative data) or other treatment for depression, particularly between 16 and 26 weeks, but it was not ethical to withhold such treatment in participants with moderate to severe depression. A meta-analysis of placebo responses in RCTs of rTMS in TRD reported response and remission rates of 20% and 11% at the end of treatment, versus 33% and 19%, respectively, in BRIGhTMIND.155 The long duration of illness and degree of treatment resistance of the participants in BRIGhTMIND also suggest that the magnitude of effects seen are unlikely to be placebo-related. In BRIGhTMIND, the mean duration was over 9 years. In terms of treatment resistance, while the magnitude of the effect of sham TMS in patients with depression can be large (g = 0.8), this drops in studies of those whose depression has failed to respond to 1 or more treatments (g = 0.67) and even further for 2 or more treatments (g = 0.5).156 In BRIGhTMIND, all had failed at least 2 and the majority at least 4.
Therefore, it is likely that the sustained improvements over 26 weeks in both treatment groups is likely to reflect a significant treatment effect. The response and remission rates in this study are striking, especially comparing them with those seen in similar patients with chronic persistent depression in specialist mental health care across three sites in England, receiving usual mental health care over 6 months (15% and 9% for response and remission respectively – less than half the response rates with TMS in this trial).58 Overall, we conclude that the clinically substantial improvements in depression over 26 weeks including sustained remission seem to be largely due to rTMS or cgiTBS alongside usual care for moderately severe depression, but some of the improvement was also likely to be due to regression to the mean and non-specific effects other than TMS.
Equality, diversity and inclusion
National and international studies report that the gender ratio (women: men) of depressive disorders is over 1.7 for lifetime prevalence, indicating that women are about twice as likely as are men to develop depression during their lifetime.157 The study population, on the other hand, had roughly equal rates of males and females, which is contrary to other French and Canadian TMS trials, which included higher rates of women than men.23,28 Therefore, it is possible that TMS may be an acceptable treatment to males in the UK. This is important given that in the UK about three quarters of deaths from suicide each year are men.158 The mean age of the present study’s participants was also 43.7 years (range: 18.0 to 83.0 years), indicating a wide spread of age, and in keeping with the rates of moderate and severe depression symptoms by age group.159
In addition, we noted that only 9% of the sample were of ethnic minority backgrounds. This is lower than expected given that prevalence of depression is reported to be higher in these groups in the UK than in the majority white ethnic group.160,161 We expected that recruitment would be challenging, given that ethnic minority groups are reported to be less likely to seek professional help for mood disorders,162 which is why we selected sites to reflect geographical diversity (London, North and Midlands), adding Oldham deliberately to improve ethnic diversity (particularly South Asian population). However, we were limited by the need to include NHS sites with experience and expertise with TMS (Nottingham, Newcastle, Camden and Islington) or other invasive treatments for TRD (Newcastle). We also ensured that the LEAP included members from different ethnic backgrounds (3 South Asian and 1 Chinese ethnicity) and included a range of age, different experience of TMS and other treatments for severe or chronic depression, included a carer and representatives with neurodiversity. They commented on all materials and procedures used in the study to ensure that they would be understandable to people from different ethnic, demographic, and neurodiverse backgrounds. Further LEAP member involvement in the BRIGhTMIND study is reported in the PPI Chapter 5.
A final limitation of the study is that the ethical requirement for capacity to consent precluded the inclusion of the most severely ill patients with depression who were unable to give consent.
Safety, acceptability, facilitators and barriers
The study demonstrates that cgiTBS and rTMS treatments are safe, but with a lot of minor side effects, which were largely resolved either the same day or within 1–6 days. There were two SAEs reported as being possibly related, including a manic episode and a psychotic episode with severe anxiety/depression. However, there were situational factors and a life event to take into consideration for these two participants, and while TMS may have contributed to these events they were unlikely to be the only contributing factor. Overall, cgiTBS was associated with more minor and moderate AEs than rTMS, including these requiring more drug treatment. The most frequently reported side effects overall were headaches, which is comparable with prior TMS trials.23 Furthermore, whilst the side effect profiles between rTMS and cgiTBS were largely comparable, there did appear to be a slightly higher proportion of participants experiencing tinnitus for the cgiTBS group, which is contrary to the THREE-D trial findings.23 All the participants were advised to wear earplugs. However, factors that can affect the amount of sound reaching patients’ ears following TMS include, but are not limited to, the duration, pattern and intensity of TMS pulses; the closeness of the coil to ear; hearing protection; and coil contact with the head.163–165 Perception of loudness also integrates over time,166 so as iTBS pulses were closer together than rTMS pulses, this may have likely lead to greater perceived loudness. The mean cgiTBS target locataion was also closer to the ears than the rTMS location, which too may have led to higher rates of tinnitus. This further emphasises the significant importance of using well-fitted, approved hearing protection to reduce the chances of temporary or permanent hearing changes, and that there needs to be greater research into side effect profiles of longer durations of TBS, multiple TBS sessions in a day (accelerated protocols) and personalised target sites. In addition, while a small proportion of participants reported cognitive difficulties such as brain fog and difficulty findings words, as well as increased anxiety, these events were largely transient, not detectable on the THINC-it battery at follow-up, with the clinical outcome data demonstrating improvements in cognition and anxiety as previously discussed.
Furthermore, while there were more mild and moderate adverse events in the cgiTBS group, there were more participants in the rTMS group that needed to reduce their percentage rMT with changes of position both temporarily and throughout the course of treatment to ensure that the TMS course was tolerable. The qualitative work also indicated tolerability of treatment improved over the course of sessions, which is in line with quantitative results reported in prior rTMS trials.167 We tested motor threshold at the first session and tested again at the sixth session. However, recent evidence suggests that percentage rMT can vary significantly across days, and weekly motor threshold testing can increase the accuracy of treatment dose determination.168 Therefore, it could be suggested that testing motor threshold weekly may help participants to achieve accurate treatment doses.
The qualitative and quantitative work also indicated that both TMS protocols are an acceptable treatment for this population in both the short term and longer term, with 92% of participants completing all 20 TMS sessions within the 4- to 6-week treatment period. The study also demonstrated that patients with TRD are particularly open to trying non-pharmaceutical treatments, with a clear understanding of the need for research in order to introduce new treatments and advance knowledge about the disorder.
Staff reassurance and communications was also an important facilitator, which improved the acceptability of treatments, with staff interactions perceived to have had therapeutic benefit, in line with prior research.55 The findings of the study clearly set the clinician as facilitators of TMS for those with TRD, requiring staff to be approachable and show empathy, as well as preparing to explain the treatment to patients and reassure them about the side effects and clarify the differences with ECT. Clinicians need to also be able to describe the TMS sensation to patients and illustrate the treatment journey while monitoring patients’ expectation of the treatment.
From a clinical perspective, there should be attempts to have flexibility around the treatment schedule. Given that the treatment commitment is intense and work commitments might be an issue for patients with more rigid work schedules, offering treatment in the evening as an alternative will prove useful for patients. Alternatively, the effects of condensed TMS treatment sessions delivered over 1 or 2 weeks might also be more feasible for the working population wanting to access TMS treatments.48,49
Finally, while some participants did report concerns that treatment felt varied dependent on the TMS staff delivering the sessions, the neuronavigation data indicated that there were only very small median differences in distance and angle between the first session and subsequent sessions. In line with prior research that shows there are still some variabilities in terms of target positioning and coil drift (coil moving during a session) for neuronavigated TMS over sessions.24 Nevertheless, these reports indicate that patients can feel even the smallest of adjustments.
Economic evaluation
The cost-effectiveness analysis of the BRIGhTMIND RCT suggests cgiTBS is a cost-effective alternative to rTMS over the 6-month trial time horizon. The analysis finds cgiTBS is associated with a 0.009 QALY gain and a £180 cost saving to the health service per individual. Cost savings and QALY gains were largest when using unadjusted (observed complete case) analyses and when considering broader societal costs. The cost-effectiveness of cgiTBS was predominantly driven by intervention-, outpatient- and productivity-related cost savings and from improvements in HRQoL relative to rTMS (primarily in usual activity, pain/discomfort and anxiety/depression health domains). Informal care hours also reduced fastest in the cgiTBS arm during the trial. Results suggested considerable decision uncertainty with the 95% CI for costs and QALYs overlapping between rTMS and cgiTBS, although the probability cgiTBS being cost-effective remained above 80% across a wide range of thresholds at base-case settings. Findings were highly sensitive to the exclusion of rTMS imaging costs, with cgiTBS having an ICER (£32,920) above NICE’s maximum threshold (£30,000/QALY) when making this cost adjustment. These findings were offset however when considering 30-minute time savings for cgiTBS treatment appointments relative to those for rTMS. The marked improvements in HRQoL observed following both treatments translated into health gains considered borderline cost-effective when assuming in the absence of treatment non-intervention healthcare costs are unchanged, HRQoL remains constant at baseline values and equipment costs are distributed across 98 or more individuals.
This is the first study to estimate the cost-effectiveness of cgiTBS relative to rTMS in the UK context. Study findings were broadly consistent with related works and other findings from the BRIGhTMIND study. Mendlowitz et al.’s cost-analysis demonstrated iTBS as a cost-saving intervention relative to rTMS within the Canadian healthcare setting.62 McCrone et al.’s UK cost analysis of TRD patients found outpatient visits and primary care contacts also constitute the highest mean health service costs.57 McCrone et al. did, however, report markedly higher inpatient costs than BRIGhTMIND.57 In line with the primary outcome analysis in this report, the economic evaluation also found (non-statistically significant) improvements in outcomes for cgiTBS that were maintained over the trial period relative to rTMS. Study findings also align with the global analysis of patients’ impressions of change during treatment, showing participants typically felt ‘somewhat’ or ‘much better’ on cgiTBS compared to rTMS.
This economic evaluation was largely confined to the scope of the BRIGhTMIND study. Future economic evaluations of electromagnetic therapies in TRD can go further by considering a broader range of comparators than those considered in BRIGhTMIND (e.g. pharmacological augmentation strategies, psychotherapeutic strategies, DBS, etc.); assessing cost-effectiveness over longer time horizons, incorporating broader forms of evidence (e.g. observational data, meta-analytic findings, expert opinion, etc.) using decision analytic modelling approaches,169 and accounting for the broader impacts TRD alleviation may have on individuals, families and wider society.
Strengths of this study include a relatively large and diverse sample of TRD patients from which costs were compiled using relevant UK costing sources and HRQoL measured using a validated instrument; study methodology was consistent with NICE methodological guidance and reviewed by a LEAP; there was a high completion rate of all economic measures at all time points; findings considered alternative analytical methods (unadjusted and adjusted approaches); cost-effectiveness was assessed using multiple costing perspectives, and the robustness in base-case findings were explored via scenario and threshold analyses. Limitations include highly uncertain predictions resulting from treatment effects estimated from a trial not powered to detect for such differences in costs and QALYs; no extrapolation of treatment benefits beyond the trial’s 6-month time horizon, and the generalisability of the findings limited by the trial’s exclusion criteria, treatment protocol (imaging for rTMS and comparable appointment times between arms), and smaller-scale equipment throughput. Furthermore, in the absence of a placebo/sham arm in the trial, assessments for the cost-effectiveness of cgiTBS and rTMS relative to no electromagnetic therapy could not be directly established. The artificial placebo scenario explored in this economic evaluation is largely hypothetical and must be interpreted with caution.
The cost-effectiveness for cgiTBS could have been underestimated for a number of reasons. First, expert opinion suggests cgiTBS sessions can provide significant time savings for healthcare professionals compared to TMS appointments.28,62 Mendlowitz et al. reports the 30-minute time saving per appointment the single-largest cost-saving factor in their analysis (US$394, 2018).62 While these direct savings were explored in scenario analyses, this economic evaluation may neglect potentially broader benefits from improvements in treatment capacity within an already highly constrained setting. Second, the removal of imaging costs considered for rTMS may also be applicable to cgiTBS given the Beam method allows neuronavigation without the need for an MRI scan. Third, HRQoL benefits associated with cgiTBS were present at final follow-up. This benefit combined with the potential for a longer duration of effect compared with rTMS, as evidenced by cgiTBS having a higher proportion of sustained responders,43 suggests a longer-term follow-up would likely capture additional benefits not accounted for within this evaluation.
Overall, evidence from the BRIGhTMIND trial suggests cgiTBS could be considered a cost-effective alternative to rTMS for patients with moderate to severe TRD. In all economic analyses, cgiTBS was considered cost-effective from a health service perspective provided it cost £184 or less than rTMS treatment (cgiTBS was a cost-saving intervention in all base-case analyses). The cost-effectiveness of cgiTBS is particularly dependent on imaging costs and the potential for larger staff cost savings than those observed in BRIGhTMIND. The overall cost-effectiveness of cgiTBS and rTMS treatments compared to non-electromagnetic treatment options is likely to be contingent on the longevity in treatment benefit, health service, societal cost savings and equipment throughput.
Future research recommendations might focus on delivering a pragmatic clinical trial to establish the clinical and cost-effectiveness of delivering accelerated brief sessions of iTBS up to five times per day versus rTMS without MRI scans within a usual care setting to reflect growing TMS practice. While the current research makes the case for both being cost-effective it required the use of applying assumptions about the care pathway that were untested because of the research nature of the project under EME guidelines for example matching of number of pulses and duration of TMS sessions. Such a trial could include the important null hypothesis or third arm of testing the effectiveness and cost-effectiveness of delivering transcranial stimulation methods (i.e. rTMS and cgiTBS) compared to routine NHS or standard care for individuals with treatment-resistant moderate to severe depression without any transcranial stimulation treatment. Further studies and models could consider the longer-term efficacy of transcranial stimulations on this patient population beyond the 6-month trial period. BRIGhTMIND showed promising results for both cgiTBS and rTMS up to 6 months but there might be further gains over routine care over 2 years or not of more.
Mechanism of action
Based on the findings and these prior MRI-neuronavigated RCTs, it could be that the method for targeting the DLPFC is perhaps less relevant, and if the DLPFC site of choice is accurately and consistently targeted during and at each session, then the stimulation of neural circuits of interest may be optimised, which may lead to more sustained depression responses. Neuronavigation is advantageous in this respect, as it has less coil drift and has significantly less off-target placement, when compared to traditional elastic cap scalp targeting.24 While one previous MDD study reported greater clinical efficacy for MRI-guided neuronavigated TMS versus scalp-based targeting methods,25 others have found no difference in clinical efficacy.26,27 Importantly, these studies focused on immediate efficacy rather than longer-term efficacy; however, future research may want to compare clinical efficacy for MRI-neuronavigated TMS versus non-neuronavigated TMS over longer-term follow-ups. This may be particularly important given the costs of additional neuronavigation systems, MRI scans, and staff to process scans and identify target co-ordinates.
On the other hand, while we cannot at present make any confident conclusions about image guidance, one of the findings in BRIGhTMIND is consistent with the hypothesis of that therapeutic effect is greater for a target site with greater effective connectivity from target to rAI, which provides modest evidence in favour of the image guidance. Specifically, the hypothesis that the baseline influence of the rAI on the lDLPFC would predict clinical response to TMS was not confirmed. However, there was evidence that the reverse (the influence of lDLPFC on rAI), and the balance of influence between the two regions, could predict improvement on the observer-rated HDRS-17. Greater influence from lDLPFC to rAI at baseline was associated with greater improvement. The rAI locus was used in pilot work based on previous observations that glucose metabolism within this region was able to distinguish responders from non-responders to antidepressant treatments.131 It is conceivable that a stronger baseline influence of lDLPFC on rAI would allow the effects of TMS to lDLPFC to spread more readily to rAI, thereby exerting greater clinical benefit. In post hoc analyses, we will therefore explore the extent to which clinical response can be predicted from proximity of the actual used target to the point within lDLPFC with maximal influence on the rAI in each participant. This may yield an alternative treatment targeting approach that could be explored in future work.
More negative baseline FC (anti-correlation) between the sgACC and the targeted region within the lDLPFC has previously been shown to predict short-term clinical improvement.134 This finding built on work showing abnormally elevated sgACC activity in depression, and reduction of sgACC activity in successful treatment response.170,171 Examining several potential lDLPFC targets, Fox et al. found that the ‘F3’ 10 : 20 scalp location, as used for the rTMS group in the current study, had the weakest anti-correlation with the sgACC.44 Nevertheless, the rTMS seed showed stronger anti-correlation with sgACC than did the cgiTBS seeds. It may be that the algorithm for identifying the cgiTBS seed inherently selects seeds with weaker anti-correlation with the sgACC, since the algorithm selects the DLPFC target with greatest (positive) effective connectivity from the rAI, and the activities of rAI and sgACC are themselves positively correlated.
In the study, FC between the sgACC and the lDLPFC target at baseline did show the expected (albeit weak) negative relationship with self-reported clinical improvement, when measured shortly after the end of the treatment course. It may be, since lDLPFC exhibits negative FC with sgACC at baseline, that ‘excitatory’ stimulation to IDLPFC gives rise to suppression of sgACC activity (and thus therapeutic effect). This suppression may be more effective when negative FC between the lDLPFC and sgACC is stronger. It is interesting then, that the relationship between baseline DLPFC-sgACC connectivity and clinical response is reversed when considering the later follow-up time points in the data (16 and 26 weeks). The effect at later time points was also shown by Hopman et al. who found that baseline FC with sgACC did not predict response immediately after the end of a treatment course, but greater (more positive) FC between sgACC and a frontal region within the executive control network did predict response at 2 months post treatment.172 Despite the promise of baseline DLPFC-sgACC FC as a predictor of TMS outcome shown in other studies, it will be important in future work to examine TMS outcomes not only in the short term but also in the longer term. In post hoc analyses, we will explore the pattern of clinical response across follow-up time points in people with more, and less, negative baseline FC between DLPFC and sgACC, to clarify this relationship further, as well as the effective connectivity from the TMS targets to sgACC to examine if this is also predictive of therapeutic effect.
One of the most consistent findings in a meta-analysis of brain connectivity in MDD was greater (positive) FC between the executive control network (e.g. DLPFC) and the DMN (e.g. DMPFC).173 These networks are typically uncorrelated, or anti-correlated, in health. Greater positive correlation could represent interference between default mode brain areas involved in internally directed mental activity and rumination and areas involved in attention and executive function. Indeed, in non-clinical populations, poorer performance, and greater performance variability, on attentional tasks is related to more positive FC between the default mode and executive control networks.174,175
We found that improvement on self-report depression measures was associated with greater reduction, from baseline to follow-up, in FC between the DLPFC and DMPFC (consistent with findings from a small sample by Liston et al.).67 This relationship was not seen on the observer reported HDRS-17 and HDRS-6 measures, suggesting that it most strongly reflects a reduction in the subjectively experienced intrusion of internally directed mental activity (e.g. rumination) during attempts to attend to the external world. Of note, the self-report measures include items on concentration or decision-making, whereas the HDRS-17 does not have such items.
From the pre-specified hypotheses, the decrease in FC between lDLPFC and lDMPFC from baseline to follow-up was the only relationship of connectivity change associated with clinical response. Separately, we plan to explore relationships between change in lDLPFC and lDMPFC connectivity (or, more broadly, change in connectivity between the default mode – DMN, and executive control – ECN, brain networks) and change in performance on the THINC-it cognitive tasks. This is under the hypothesis that abnormally elevated DMN-ECN connectivity in MDD may reflect intrusion of internal-world processing (e.g. rumination) on external world processing. MDD can be associated with marked cognitive difficulties,176,177 and such difficulties may be primary mediators of occupational and psychosocial dysfunction.178 This connectivity relationship could represent a treatment target for cognitive difficulties in MDD. Taken together the three findings of associations between clinical outcome and effective or FC could be seen as different facets of a more homogenous improvement from TMS or they could represent different biotypes of response, possibly related to the course of improvement of TMS. However, there is no evidence in any of the findings that cgiTBS or rTMS are inherently associated with different patterns of improvement or clinical response.
Additionally, the spectroscopy findings add to the plethora of inconsistent pre-trial literature evidence on GABA and Glx. In line with the pilot study results,68 we did find significant associations between baseline GABA and clinical improvement, albeit in the opposite direction to what we hypothesised with the significant findings driven by the GE scanners. The findings suggest that participants with higher inhibitory tone had better response to TMS treatments, which is in line with a recent study that found greater baseline cortical inhibition was associated with better response to TMS in youth depression.179 While caution has to be applied due to the large voxel size, variability induced by different scanners and substantial quality-related attrition, this finding together with the eFC prediction may suggest that greater inhibitory influence of the dlPFC over RAI as mechanism of TMS response. Furthermore, while some studies have found increases in glutamatergic compounds are associated with better symptom improvement following TMS,75 we found decreases in Glx predicted better response on the HDRS-6 at 26 weeks (changing to trend-level significance when removing the relatively extreme value of Glx change). The findings are in line with a prior study which found decreased Glx was associated with better treatment response following CBT,180 and may reflect a regional normalisation in the prefrontal cortex (reduced hyperactivity). However, we interpret this finding with caution considering the smaller sample size within the analysis and the one relatively extreme value of Glx change.
Overall the MRS findings within the study are not robust. A major limitation includes losing a considerable number of MRS data due to QC issues, so findings are based on relatively small sample sizes. Thus, we suggest the findings are exploratory requiring the need for further replication and study. A further limitation was the use of several MRS platforms. Descriptive statistics revealed variability in signal to noise ratio across scanners, better GABA fitting error for GE scanners, but better Glx fitting error for Phillips scanners. In addition, a number of MRS scans were excluded due to miss-matched voxel locations between baseline and follow-up scans. In future studies, higher field strength (i.e. 7 tesla) scanning platforms may improve quantification of GABA and glutamatergic compounds.181
Finally, the mechanism-of-action findings refers to pre-specified analyses of the BRIGhTMIND data, selected to address the mechanistic hypotheses that motivated the trial itself. Multivariate machine-learning algorithms that draw on a large array of imaging features show promise for predicting outcomes in MDD.182 The well-characterised BRIGhTMIND dataset, with outcomes at multiple time points, variability in targeted location for cgiTBS, and imaging data from multiple modalities (including MRS), provides an ongoing resource for identifying and evaluating cutting-edge approaches for predicting treatment outcomes and for identifying potential novel treatment targets.
Summary of limitations of the trial
There was no sham or treatment as usual control group. We therefore cannot be sure that the changes in the trial were not due to non-specific effects or regression to the mean. Treatments such as TMS that are not usually available to participants may carry a high degree of hope and expectancy effects as illustrated by the qualitative data. Getting up each day for 4 to 6 weeks may have created order and routine which might have had a therapeutic effect as described in the qualitative data. Ideally, additional treatments would not be altered throughout the course of the follow-up period as well as the TMS treatment itself, but this was not ethically or clinically indicated in such a severe clinically vulnerable effect. It is therefore important to consider that this trial is of two different courses of TMS in addition to usual care. It is possible that some of the changes in outcome in both groups, especially at 16 and 26 weeks, were due to these additional treatments. Of note the qualitative data reported that they experienced an increase in anxiety after both forms of TMS but there was no evidence of a rise in GAD-7 scores at 8, 16 or 26 weeks.
Nevertheless, these are participants with a very long duration of their current depression episode, with a median duration of over 6 years and a history of treatment resistance to multiple treatment attempts. Two thirds of the sample had four or more adequate levels of treatment of sufficient duration and dosage of antidepressants, augmentation, or ECT. Moreover, this is likely to be an underestimate of their treatment resistance since psychological treatments and social interventions were not recorded. Given the duration, severity, and history of treatment resistance, regression to the mean and non-specific effects of treatment may have been quite low. Furthermore, there were substantial mean changes of clinical importance on all measures of depression, anxiety, function, and quality of life, indicating that both groups did show considerable improvement in the trial. The findings of response in a third and sustained response or remission in a fifth of these participants is therefore clinically important, even if some of these results were due to changes in usual care, changes in routine or other non-specific effects of treatment.
In terms of the sample, there were slightly more men, married/cohabiting and employed participants than might be anticipated in community samples of people with depression with few participants of retirement age. This may reflect the novelty and appeal of TMS in terms of men and participants of working age, the need to find alternative treatments to remain in employment, and the wish for someone to support the participant during a TMS course. Analysis of demographic and clinical moderators of outcome such as age, gender, and comorbidity suggest that the results are likely to be generalisable despite some of these demographic differences.
In terms of delivery of TMS, there were some errors in the correct location of the site for the TMS treatment as none of the sites had previous experience with MRI-neuronavigated TMS. However, the data on the closeness of the site of TMS stimulation and how little this varied across 20 TMS sessions suggests that this factor had little bearing on the main findings of the study. More importantly, some participants in both groups started to improve only towards the end of the course of TMS, so greater improvements in outcome might have been achieved with longer courses of TMS. In the cgiTBS group, iTBS was delivered with only 5-minute intervals between bursts of 600 pulses; optimal delivery might have been achieved with 15-minute intervals between bursts of 600 pulses.
The trial was conducted during the COVID-19 pandemic, but analysis suggested that this had little overall effect on the results, except possibly in relation to THINC-it and imaging, where there was some loss of follow-up scans and THINC-it assessments during the period of COVID-19 closedown and restrictions. Participants who completed the trial before the pandemic closedown completed THINC-it assessments at baseline, 8, 16 and 26 weeks and on restart at baseline and 16 weeks. Performance may have improved with greater practice. We also changed the primary outcome from response at 16 weeks to mean change in HDRS-17 at 8, 16 and 26 weeks. Inspection of response data shows that if recruitment to the original sample size and primary outcome had been possible, the overall results and conclusions of the study would not have been different.
In terms of bias, there were few unblindings and on few occasions did outcome assessors or participants identified their treatment allocation. Inter-rater reliability was high in terms of assessment of the primary outcome, suggesting that the results were not due to a lack of imprecision in the assessment of outcome. There were lots of major deviations of the study, but these were largely because the majority of participants undertook a break in TMS treatment of 4 days or more. Analysis of the primary outcome using various forms of sensitivity analysis showed very little difference to the ITT analysis with a consistent pattern of results across all outcomes. Therefore, the results of the study are likely to be sound.
Conclusion
Connectivity-guided intermittent theta burst stimulation was not superior in clinical efficacy to rTMS, although it was dominant in terms of cost-effectiveness but not under all assumptions. About one fifth of participants with moderate to severe TRD achieved remission or a sustained response, which raises the possibility that some patients unresponsive to their treatments could be kept well with 1 or 2 MRI navigated courses of 20 or sometimes 30 iTBS or rTMS sessions. Not all the improvement is likely to be due to the direct physiological effects of TMS but due to non-specific factors and usual care for depression. Adverse events are common but minor and time-limited, and few that can be related to TMS are serious. There are a number of changes in FC and cognition that were found, largely consistent with other emerging literature, but worthy of further exploration in this large, well-characterised sample.