Sleep as a Predictor of Improved Response to Transcranial Magnetic Stimulation for Depression (SPIRiTeD)

INTRODUCTION

Major Depressive disorder (MDD) is a common problem among the general population and Veterans are disproportionately affected (Hasin et al., 2005; Liu et al., 2019). The majority of treatments for MDD commonly include psychopharmacology and psychotherapy. Despite these treatments, only about a third of patients achieve remission and many demonstrate a response to treatment (Gaynes, 2016; Pigott, 2015; Yesavage et al., 2018). Transcranial magnetic stimulation (TMS) and electroconvulsive therapy (ECT) are interventional treatments typically reserved for treatment-resistant MDD (TRD).

TMS was cleared by the U.S. Food and Drug Administration (FDA) in 2008 for treatment of MDD and has been shown to be an efficacious treatment for MDD even in the setting of comorbid psychiatric disorders (Hernandez et al., 2021; Thompson, 2020). TMS for MDD is a noninvasive treatment that involves utilizing electromagnetic energy to stimulate the left dorsolateral prefrontal cortex (DLPFC) to alter cortical excitability and corresponding neurocircuits. The original FDA-cleared TMS protocol for MDD involves stimulation over this brain region at a frequency of 10 hertz (Hz) for 3000 pulses at 120% of the individual’s resting motor threshold (rMT) (George et al., 2010). Treatments are typically performed once a day, 5 days per week for approximately 6 weeks. Despite encouraging results from TMS treatment, many patients still do not achieve remission with current TMS treatment protocols (Kozel et al., 2017; Madore et al., 2022).

Efforts are ongoing to optimize TMS treatment. Alterations to TMS parameters have led to additional FDA-cleared protocols for the treatment of MDD including utilizing different pulse frequencies such as intermittent theta bust (iTBS) (Blumberger et al., 2018), shortening of time between pulse delivery (known as the interstimulus interval) with 10Hz stimulation (Carpenter et al., 2021), and delivery of multiple sessions per day often referred to as accelerated protocols (Cole et al., 2020; Zuschlag et al., 2023). Additional modification and/or combination of stimulation parameters such as frequency, laterality of electromagnet placement, and coil configuration have been studied to optimize treatment response and remission rates (Li et al., 2021). Adjunctive techniques to improve delivery of treatment have been investigated as well, including neuronavigation, such as MRI-based 3-D modeling and resting state fMRI proxy measures of connectivity to ensure accurate and consistent targeting of underlying brain structures, and EEG for measuring electrical response (Lioumis and Rosanova, 2022; Tremblay et al., 2019).

Furthermore, four trajectories of response to TMS have been identified that do not appear to be impacted by TMS protocol (such as neuronavigation with accelerated TMS, iTBS or traditional 10Hz TMS)(Kaster et al., 2020; Kaster et al., 2019). Trajectories based on response, as identified by Kaster et al, included “nonresponse” (minimal response throughout treatment), “rapid response” (maximum response attained by week 2–3), “higher baseline symptoms, linear response” and “lower baseline symptoms, linear response” (both with linear improvement throughout treatment, separated by baseline symptom severity). This suggests that undefined factors contribute to TMS response. Broad categories have been proposed which include patient related factors, illness related factors and TMS procedural factors (Kar, 2019). More sophisticated biomarkers are also being investigated for predicting response such as motor evoked potential (MEP) and EEG (Philip et al., 2019; Rocchi et al., 2018). Continued investigation of these factors and their impact on TMS treatment can guide future efforts to optimizing TMS treatment.

One clinically relevant aspect of MDD is sleep dysfunction. Reports show that as many as 90% of patients with MDD report sleep dysfunction and as many as 40% meet criteria for insomnia disorders (Franzen and Buysse, 2008; Stewart et al., 2006). Sleep dysfunction is associated with poor outcomes in the setting of MDD (Cutler, 2016; Mason and Harvey, 2014; Pigeon et al., 2012; Soehner et al., 2014). Baseline sleep dysfunction may improve with TMS treatment for MDD (Hines et al., 2021) and has also been suggested to be a positive predictor of the antidepressant effects of TMS (Brakemeier et al., 2007). Yet not all literature supports this positive association. For example, one study examining sleep and TMS for depression found no significant difference on sleep related mental health assessment scores between active and sham treatment groups (Rosenquist et al., 2013). Another study found improvements in depression and sleep dysfunction with TMS treatment, however improvements of sleep and depression rating scales were independent of each other and did not correlate (Collins et al., 2022; Kaster et al., 2023). The relationship between sleep and depression during TMS treatment has not been fully characterized, and more information on the relationship between sleep and depression response is needed (Centorino et al., 2020). This warrants further investigation as sleep dysfunction is a factor that can be assessed for and possibly addressed early in TMS treatment. Further, a small proof of concept study has shown that implementing a treatment for comorbid Insomnia Disorder, specifically cognitive behavioral therapy for insomnia (CBTi) while treating MDD with rTMS, is feasible and may have a beneficial effect (Norred et al., 2021). This study aimed to address the current gap in the literature by investigating if improvement in baseline sleep dysfunction is a predictor of improvement in depression symptoms during TMS treatment.

METHODS

A retrospective observational cohort study was conducted examining Veterans receiving TMS treatments through the U.S. Department of Veterans Affairs “VA TMS Pilot Program” from March 2017 through March 2020 (i.e., Veterans completing treatment prior to the start of the COVID-19 pandemic). The VA Palo Alto Healthcare System Mental Illness Research Education and Clinical Center (MIRECC) served as the coordinating site for the study, which was approved as quality improvement by the regulatory bodies of VA Palo Alto and the Stanford Institutional Review Board. The “VA TMS Pilot Program” is a multi-site clinical, educational, and training program implemented to distribute TMS devices throughout the Veterans Health Administration (VHA) and provide infrastructure for optimizing the use of TMS within VHA (Madore et al., 2022). Data for this analysis was contributed by individual sites and gathered on VA REDCap for the national quality improvement project.

Veterans included in the database were referred by their primary mental healthcare provider for TMS treatment for MDD. All referred Veterans were evaluated for appropriateness of TMS treatment by a TMS-credentialed VA physician. Standard indication criteria were used to determine the appropriateness of treatment, including determination of a diagnosis of medication resistant MDD (e.g., unipolar depression, failure of >1 antidepressant in the current depressive episode). Standard TMS contraindications were utilized, such as a history of epilepsy and the presence of intracranial metal. Psychiatric medications were continued throughout TMS with no specific pre-treatment pharmacological changes required. Clinical factors, including medication adherence and substance use, were assessed daily as per standard clinical care. All Veterans in the database were included as part of this study.

The majority of TMS treatments were delivered using Magstim devices (i.e., Horizon Lite, Horizon Performance, and Rapid ² systems (Wales, UK).) Additional devices less commonly used included Neuronetics (Malvern PA, USA), MagVenture (Farum, Denmark), and Brainsway (Jerusalem, Israel) systems. Sites were advised to follow FDA guidelines, although parameters could be modified to evidence-based off-label approaches based on clinical judgment. The most used protocol was the FDA-cleared high-frequency stimulation protocol consisting of 10Hz stimulation applied to the left DLPFC at 120% of an individual’s resting motor threshold administered once daily for a period of 30 treatments, followed by a six-session taper. The most common cortical mapping and MT determination techniques included localization of the left DLPFC utilizing the Beam F3 method and MT determination via visualization of elicited motor activity in the right upper hand (i.e., visible muscle contraction of the abductor pollicis brevis and/or dorsal interosseous) with the resultant MT being the amount of energy applied to the motor cortex to elicit contralateral finger movements 50% of the time.

Mental health assessment scales were conducted and used for symptomatic monitoring for clinical purposes, including the Patient Health Questionnaire 9-item scale (PHQ-9) (Kroenke et al., 2001). These scales were obtained at baseline and after every 5^th treatment including the final day of the treatment course. Sleep-specific assessments scales were not conducted systematically across sites and thus were not available in the utilized database. The PHQ-9 was most consistently utilized rating scale used across all sites that included a sleep assessment. Thus individual items on the PHQ-9 were used to assess self-reported sleep dysfunction at baseline, as well as to assess sleep improvements during TMS treatment. PHQ-9 item number 3 (PHQ9 I3) was the item primarily used to define sleep dysfunction. Item number 3 requests the patient to rate the frequency of trouble falling or staying asleep or sleeping too much, with 0 (i.e., PHQ I3=0) being “not at all” to 3 (i.e., PHQ I3=3) being “nearly every day.” A score greater than 0 qualified as sleep dysfunction and a reduction in the rating score from baseline was considered sleep improvement.

Associations between improvements in sleep dysfunction and improvements in depressive symptoms with TMS treatments were analyzed at multiple time points including at the end of week 1, week 3, and week 6 (end of treatment): defined as early, middle, and late improvement of sleep dysfunction, respectively. Clinically significant improvements for depression were defined per the standards of the respective assessment scales with remission defined as less than 5 on the PHQ-9. When examining depressive symptom improvement via a reduction in total PHQ score, a modified PHQ-9 was utilized with the sleep item removed to avoid bias, resulting in a PHQ-8.

Descriptive and inferential statistics were conducted using IBM SPSS Statistics version 28 (IBM Corp. Released 2021. IBM SPSS Statistics for Windows, Version 28.0. Armonk, NY: IBM Corp). Chi-square tests were utilized to examine differences in baseline/pre-treatment sleep dysfunction and end of treatment depression remission rates. Pairwise comparisons with a Bonferroni correction were utilized to examine the differences between severity of baseline sleep dysfunction and depression remission rates. Further analyses examined the relationship between improvements in sleep dysfunction at early, mid, and late time points to end of treatment depression remission rates. Univariate analysis of variance was used to examine improvements in sleep dysfunction in relation to final PHQ-8 scores.

RESULTS

A total of N=854 Veterans were treated across 27 VAMC sites. Demographics, baseline patient characteristics, and treatment parameter details can be found in the original report examining depression response in this cohort (Madore et al., 2022).

Baseline Sleep Dysfunction:

Of the N=854 identified Veterans, N=825 had available baseline sleep data, as represented by the sleep item of the PHQ-9. Of these, 94.30% (N=778) reported sleep dysfunction at baseline, with the highest severity of dysfunction on the assessment scale (e.g., PHQ-9 I3= 3; experiencing sleep dysfunction “nearly every day”) being the most commonly reported sleep dysfunction (59.5%, N=490). Initial Chi-square test of independence was run to examine the relationship between baseline severity of sleep dysfunction and post-treatment remission. The test was significant, χ2(3,N=541)=23.71,p<.001, indicating a significant association between sleep dysfunction and remission. The effect size was small (Cramér’s V = 0.21). The observed frequencies were as follows for Veterans not in remission based on PHQ-9 Item 3 answers are as follows: 16 for (0) “Not at all”, 53 for (1) “Several days”, 92 for (2) “More than several days”, and 260 for (3) “Nearly every day”. The observed frequencies for Veterans who met remission criteria based on PHQ-9 Item 3 answers are as follows: 14 for (0) “Not at all”, 20 for (1) “Several days”, 39 for (2) “More than several days”, and 48 for (3) “Nearly every day”. Posthoc analysis for all pairwise comparisons for Veterans in remission was then conducted to further investigate between group differences. Utilizing Bonferroni corrected values, the results indicated that a higher number of individuals experiencing the most severe level of baseline sleep dysfunction met end of treatment depression remission criteria based on PHQ-9 total scores, however, rates of remission were lower in the group with highest severity sleep dysfunction.Statistically significant difference were seen between the highest severity of baseline sleep dysfunction and differences in end of treatment depression remission rates when compared to those without reported sleep dysfunction (p=.001) and compared to those with “moderate” level sleep dysfunction (p=.007). No other between group comparisons were significantly different.

Improvements in Sleep Dysfunction

Further analyses were conducted examining the relationship between improvements in sleep dysfunction during TMS treatments and depression symptom improvement. Of note, due to the nature of clinical data, the n analyzed for each time point was dependent on the number of completed assessments. 37.38% (194 of 519) of Veterans receiving TMS for MDD had improvement of sleep dysfunction at the end of week 1 of treatment, 51.75% (266/514) had improvement of sleep dysfunction at week 3, and 57.0% (213/374) had improvement of sleep dysfunction at week 6. Results of McNemar chi-square analysis demonstrated an association between sleep improvements and depression remission rates, and Veterans with improvement of sleep dysfunction were observed to have higher rates of depression remission at completion of TMS treatment. This was observed to be statistically significant for all timepoints analyzed: week 1, 27.84% vs 18.15% (p= <.001); week 3, 29.32% vs 14.52% (p= <.001); week 6, 30.99% vs. 9.94% (x2 [1, N = 374] = 23.730, p= < .001). Univariate analysis of variance examining sleep improvements and final PHQ total score (modified to a PHQ-8 as noted above) found greater improvements in end of treatment PHQ-8 scores for those with sleep improvements compared to those without sleep improvements. Again, this was observed and was statistically significant across all three time points, with differences in mean improvements in final PHQ-8 scores showing: week 1 (M = 2.438, SE = 0.625, 95% CI 1.211 to 3.666); week 3 (M = 3.211, SE = 0.603, 95% CI 2.025 to 4.396); and week 6 (M = 4.825, SE = 0.689, 95% CI 3.471 to 6.180.) Results are summarized in Table 1.

Table 1:

Summary Results Showing the Association Between Sleep Dysfunction Improvement and End of Treatment Depression Symptomatology. Veterans who experienced sleep improvements had higher end of treatment depression remission rates and greater reductions in end of treatment PHQ total scores. This was statistically significant at all three time points analyzed. McNemar’s chi square test was used for remission rate analysis and ANOVA for PHQ total score analysis.

Association Between Sleep Dysfunction Improvement and End of Treatment Depression Symptomatology
	Treatment Week
Depression remission rates^		Sleep Improvement	No Sleep Improvement	p
	Week 1	27.84%	18.15%	<.001*
	Week 3	29.32%	14.52%	<.001*
	Week 6	30.99%	9.94%	<.001*
Final PHQ scores&		Differences in Mean Improvements	SE	95% CI
	Week 1	2.438	0.625	1.211 to 3.666*
	Week 3	3.211	0.603	2.025 to 4.396*
	Week 6	4.825	0.689	3.471 to 6.180*

^{^}Remission was defined as a PHQ-9 total score of less than 5.

^&Final PHQ total score was modified to a PHQ-8, removing the sleep item to avoid bias.

PHQ= Patient Health Questionnaire

DISCUSSION

To the authors’ knowledge, this study has the largest sample size to date looking at the relationship between improvement in sleep dysfunction and TMS treatment for MDD. While previous data has been mixed regarding sleep improvement as a predictor of TMS response for MDD, our study suggests improvements in sleep dysfunction are associated with higher rates of end of treatment depression remission and lower PHQ-8 scores.

The authors hypothesized that early improvement of sleep dysfunction would be a clinical predictor of improvement in depressive symptoms during TMS treatment for MDD. In support of our initial hypothesis, improvement of sleep dysfunction at weeks 1, 3 and 6 were all found to be associated with higher end of treatment depression remission rates and greater reductions in final PHQ-8 total scores (modified PHQ-9 by removing the sleep item).

When looking at baseline sleep dysfunction, lower rates of depression remission were generally seen with increasing severity of baseline sleep dysfunction. Each time point showed a statistically significant difference in depression remission rates between those with improvement of sleep dysfunction and those without improvement. A notable finding was that throughout the three identified time points, the end of treatment depression remission rates in those with improvement of sleep dysfunction did not change substantially (27.84%. 29.32%, 30.99% at weeks 1,3 and 6, respectively), however, there was a relatively notable decrease in end of treatment depression remission rates in those who did not have improvement in sleep dysfunction (18.15%, 14.52% and 9.94% at weeks 1,3 and 6 respectively). This may suggest an association between persistent sleep dysfunction throughout treatment and reduced likelihood of depression remission.

Additional aims of this study included identifying rates of baseline sleep dysfunction and examining the relationship between baseline sleep characteristics and sleep improvement during TMS treatment for MDD. The significant majority (94.30%) of individuals reported baseline sleep dysfunction. Improvements in sleep dysfunction were observed with TMS treatment for MDD, with approximately one-third of Veterans having sleep improvement from baseline to week 1. Higher rates of sleep improvement were seen at later weeks, as over half of Veterans saw improvements in sleep at week 3 and week 6.

Our results show that improvements in sleep dysfunction are associated with improved outcomes with TMS treatment for MDD and supports the notion that sleep disturbance in depression can vary in intensity and type and can improve (or not improve) independent of changes in other depressive symptoms. While sleep dysfunction is a core symptom of MDD, sleep dysfunction and depressive disorders exist as parallel processes, where an improvement in one does not necessarily equate to an improvement in the other. Although the causative nature of improved sleep leading to higher remission rates cannot be determined with this study design, this study does highlight a reciprocal impact of changes of sleep dysfunction and depressive symptoms. This supports the possibility that targeting both sleep dysfunction and depression may further improve outcomes, specifically that interventions to improve sleep dysfunction during TMS treatments may be of value in optimizing TMS treatments for MDD. Future research is critically needed to investigate whether additional interventions specifically targeted to improve sleep will improve clinical outcome of TMS for MDD. Additional studies may further our understanding by utilizing more sophisticated sleep measures including sleep-specific mental health assessment scales (e.g.. Insomnia Severity Index, Pittsburgh Sleep Questionnaire), sleep diary data or more objective sleep measures, such as actigraphy, and should control for psychiatric comorbidities and primary sleep disorders. Furthermore, future work could focus on interventions to improve sleep dysfunction as a potential mechanism for increasing depression remission during TMS treatment courses for MDD in those with sleep dysfunction, such as CBT-I or psychopharmacological sleep aids.

Limitations

Given the retrospective nature, this study was limited by the measures that were previously collected as part of standard clinical care. Since no formalized sleep assessment scales were utilized across sites, this study used the sleep item on the PHQ-9 as the primary means to assess sleep dysfunction and sleep change, and thus a broader proxy for sleep dysfunction. The sleep item is a subjective, self-reported assessment of sleep where patients can rate the impact of their sleep dysfunction on a scale of 0 to 3. This limits the ability to detect change in sleep, as well as interpret the magnitude of sleep changes. Using the single item also does not provide the required information to identify whether the patient had clinical insomnia or hypersomnia. Additional limitations related to the retrospective nature of this study include the lack of ability to control for baseline differences (e.g., comorbid medical and psychiatric disorders, psychiatric medications (including sleep aids), or primary sleep disorder such as obstructive sleep apnea, narcolepsy, restless leg syndrome).

Conclusion

In conclusion, the results support our hypothesis that sleep improvements have a predictive quality on the antidepressant effects of TMS. Veterans with improvement of sleep dysfunction were found to have higher end of treatment depression remission rates than those who do not experience sleep improvements. Additionally, the finding of lower remission rates in the absence of sleep improvement may suggest guided efforts to improve sleep can improve depression outcomes in TMS. Although these findings are of clinical interest, additional research is needed to further characterize the relationship between sleep dysfunction and MDD response to TMS treatment.

Determining which patients will respond to TMS treatment for MDD is clinically important, as clinical predictors can help to identify methods to modify or augment treatment that can be implemented early and improve TMS treatment outcomes. This study supports the value of efforts toward monitoring and improving sleep dysfunction throughout TMS treatment for MDD.