In this study we have demonstrated that, rather than all symptoms showing a uniform change over time, there seems to be a temporal order in symptom change that is predictive of treatment outcome and different in those receiving active and sham SNT, suggesting that this difference is caused by treatment effects. More specifically, we observed that the earliest item to show differences between the active and sham groups is MADRS item 7 (lassitude), despite this symptom not being significantly different between groups at baseline. Lassitude showed an early reduction in the active group and remained significantly different between groups from day 2 of treatment until day 5. These findings are interesting in light of recent work that has demonstrated that rTMS may indeed have differential effects in different symptom dimensions21.
Lassitude, as defined in the MADRS, is “a difficulty getting started, or slowness initiating and performing everyday activities”22. As such, it is related to, but also distinct from, other commonly reported symptoms of MDD such as poor sleep, psychomotor retardation, low energy or anhedonia, all of which may also affect the ability to carry out tasks. Indeed, in related work authors have argued that perceived energy and fatigue are related but clinically and neurobiologically distinct states23, supporting a fine-grained approach to symptom interpretation when considering the mechanistic meaning of changes in specific symptoms. The significance of this early difference in lassitude, mechanistically, may be related to how it changes the experience of patients outside of treatment- those patients with decreased lassitude may start re-engaging in their daily routines, which in turn would have a therapeutic effect similar to that seen in the behavioral activation literature, wherein changes in behavior engenders later change in mood and other depressive symptoms24,25. We further note that differences in lassitude appeared between the active and sham groups prior to differences in sleep in our sample, suggesting that there is an increased sense of being able to complete daily tasks prior to any benefits to energy which would be conferred by improved sleep.
Evidence for the mechanistic relevance of early lower lassitude scores is present. Firstly, lassitude score at day 3 predicted treatment outcome across groups in terms of immediate response and response at four weeks post-treatment, even after controlling for baseline depression severity. In addition, 7 of 9 SNT responders immediately after treatment had early changes in lassitude, compared to only 1 of 3 SNT non-responders. As such, it seems reasonable to adopt as a hypothesis for future study that changes in lassitude may be an important part of the mechanism underlying treatment response in SNT. One alternative hypothesis which should be examined in a larger study is that early improvement in lassitude may be a nonspecific marker of improvement: of the 3 participants in the sham group who remitted at any time, all 3 experienced early changes in lassitude.
The question remains as to why SNT engenders this change in lassitude. One possibility is that early change in lassitude is a feature of any effective antidepressant treatment. While significant evidence exists that early changes in total symptom severity can predict treatment response26, there is relatively little evidence from treatment studies for the importance of early changes in specific symptoms. One study found that patients who remitted when treated with hypericum or fluoxetine had greater early reductions in the general somatic symptoms question of the Hamilton Depression Rating Scale (HAM-D)27. While not precisely the same as lassitude, the somatic symptoms question on the HAM-D does include the related notions of “loss of energy and fatigability” (Hamilton, 1960). A later analysis pooling a larger dataset from the same study found that remission in the whole sample was still associated with changes in general somatic symptoms; furthermore, when considering the treatment branches separately, general somatic symptom improvement was associated with placebo group remission28. As such it is possible that early changes in lassitude are a feature of effective antidepressant treatment in general, and potentially a key “nexus” point through which effective treatments act. However, SNT does produce improvement in depression symptoms more rapidly than other treatments, and in patients for whom other treatments have not been effective. As such, SNT may achieve improvement in lassitude more quickly due to its accelerated nature, or more effectively, due to its more precise imaging-based targeting; analyses of studies of SNT compared to other treatments with sufficiently dense early symptom sampling could help to resolve this question. Another possibility is that improved regulation of sgACC function, a downstream target of SNT, is responsible for early changes in lassitude. Indeed, one recent study found that the sgACC mediates the relationship between non-exercise activity (e.g., engaging in one’s daily routine) and perceived energy (which is likely related to lassitude)29. Another (not mutually exclusive) possibility is that the primary target (LDLPFC) is responsible for reductions in lassitude. Indeed, one study found that participants asked to perform a tiring finger tapping experiment and who were exposed to a static magnetic field over the LDLPFC had reduced slowing of their tapping rate30. In addition, a study using another non-invasive brain stimulation technique, transcranial direct current stimulation, in patients with multiple sclerosis found that LDLPFC treatment ameliorated fatigue, which is related (but not identical) to lassitude31. In addition, the DLPFC has been linked to integration of information about reward and to the driving of behavior through influence over mesocortical and mesolimbic dopaminergic systems32. As such there is evidence that both LDLPFC and sgACC may be relevant to circuits which regulate energy perception and motivated behavior, both of which are clearly related to the concept of lassitude. Another hypothesis worth considering in future work would be whether SNT belongs to a category of treatments with differential impact on the symptom dimension which includes lassitude and psychomotor retardation. Indeed, in ref. 33 the authors found that escitalopram improved an “observed mood” dimension of symptoms (which included clinician-rated depressed mood, anxiety, activity, and psychomotor changes) more than nortriptyline. As such, future work could seek to determine if different categories of treatments act through specific mechanisms, one of which may involve the improvement of lassitude/psychomotor retardation. SNT may be an example of this category of treatments, and may produce rapid improvement because it preferentially modulates neural circuits relevant to these symptoms as described above.
While the results presented here suggest that SNT may have an early effect on lassitude which leads to symptomatic improvement, it is important to note that some sham responders also experienced improvements in lassitude. Part of the reason for this may be the intensive nature of the SNT treatment, as it requires patients to present early to the clinic, experience multiple treatment sessions per day, and interact with study staff and clinicians. This experience may have an activating effect which may directly reduce patient perception of their own lassitude. Further studies including larger sham SNT arms could compare sham SNT to other treatments (e.g. treatment as usual) and may help to identify if, in a given timespan, the SNT process itself, regardless of stimulation, leads to more improvement in lassitude than other treatments.
rTMS does not only produce localized effects; rather, by stimulating one region it can have effects on interconnected brain networks34. As such, we analyzed functional connectivity between the sgACC, which is both a downstream target of SNT and a hub for networks like the default mode network35, and other regions associated with known brain networks. The first analysis considered pre-treatment FC, and we demonstrated that participants with low day 3 lassitude scores showed reduced FC between sgACC and regions in the VN and DAN, and increased FC with PFC7 in the DMN. Increased deactivation in the DMN during emotion processing has been shown to be predictive of antidepressant response36; given that SNT is targeted at LDLPFC regions which are anticorrelated with sgACC, it is possible that those whose sgACC is positively connected with the DMN at baseline are those who will benefit most from SNT. Hypoconnectivity of the DAN with frontal regions has also been found in MDD; as the DAN is involved in orienting to external stimuli, Improvements in DAN function may help people attend to the daily tasks they must accomplish, and in this way reduce lassitude37. VN hypoconnectivity is also seen in MDD and has been implicated in symptoms such as psychomotor retardation, which may again be relevant to lassitude38.
The second FC analysis was focused on FC changes between pre- and post-treatment. Participants with lower lassitude scores early in treatment showed increased FC between sgACC and regions in the CEN and reduced FC with a region in the SMN. Dysfunctional connectivity of the CEN has been implicated in MDD, with the assertion that its dysfunction negatively impacts goal directed behavior39; our results may suggest improved communication of the CEN with other brain regions, resulting in improvements in goal-directed behavior- which could reasonably be expected to manifest as reductions in lassitude. The SMN has also been implicated in MDD40, though increased FC between SMN and the fronto-parietal network (FPN) was seen in escitalopram responders41. As we only examined connectivity with sgACC, more complex connectivity patterns (e.g. with FPN) may explain this finding and could be examined in a larger sample.
If replicated, these FC patterns could serve as biomarkers to help identify which patients are most likely to experience changes in lassitude and treatment response, as well as to monitor response to treatment. Asking patients about changes in lassitude early in treatment may, if replicated, be proven a useful clinical marker of treatment response in accelerated rTMS, with patients who do not have early change in lassitude potentially being candidates for treatment modification. In addition, there may be synergistic benefits of combining rTMS with behavioral activation; indeed, studies have determined that combining rTMS and psychotherapy, including behavioral activation, is feasible, but there is a lack of evidence proving efficacy in randomized trials42,43.
Our most significant limitation is the small sample size which leads to our analyses being underpowered. Indeed, the preliminary results presented here would not survive correction for multiple comparisons, though we were able to partially replicate our clinical results in a previous open label dataset. As such, this work should be considered an exploratory, hypothesis generating analysis, the results of which must be replicated in subsequent studies. Future work could also expand on these results (e.g., in order to probe group by time interactions as well as to examine mediation and moderation). The use of only one assessment of symptom burden- the MADRS- is another limitation. Other scales capture other symptoms, and in larger samples with a more complete set of daily measures, additional early symptom differences may become apparent. Future work could also use questionnaires specifically adapted for more frequent administration (e.g. ref. 44). In addition, changes in other symptom categories not often measured by questionnaires but by tasks, such as cognition, may also be relevant early predictors of treatment response, and these are not considered here beyond the MADRS question on concentration45. Another limitation is our decision to limit our imaging analysis to only those pairs including the sgACC. While this was necessary given the small sample size of the study, it does potentially mean that some relevant FC pairs have been omitted from this analysis. In addition, due to the small sample size and exploratory nature of the analysis, we did not control for potentially important covariates (e.g., age, sex, and degree of treatment resistance) which will be important in future analyses. Better powered studies in the future should therefore examine a larger number of FC pairs as well as other imaging metrics.
In conclusion, we have identified that MADRS item 7 (lassitude) is the earliest symptom which differs between patients receiving SNT and sham treatment, as early as treatment day 2. In addition, the level of lassitude at treatment day 3 is a good predictor of treatment outcome, even when controlling for baseline depression severity. Resting state functional connectivity analyses suggest that connections between the sgACC, the ultimate target of LDLPFC-rTMS, and elements of the VN, DAN and DMN at baseline may predict lassitude at treatment day 3. In addition, changes in FC between sgACC and elements of the CEN and SMN may reflect early lassitude score and may represent biomarkers of response of this symptom to treatment. This exploratory analysis provides hypotheses which should be investigated in larger datasets.