Equal remission rates and reduced length of hospital stay with twice-daily repetitive transcranial magnetic stimulation (rTMS) for major depression – A large naturalistic retrospective cohort association study

Objectives: Repetitive transcranial magnetic stimulation (rTMS) is a level 1a evidence-based treatment for major depression, but high cost of care and limited effectiveness in naturalistic cohorts have been lingering criticisms. This naturalistic, retrospective cohort analysis compares the effect of once and twice daily treatment protocols of rTMS using quality assurance data collected at an Australian private psychiatric hospital. Methods: A total of 210 inpatients self-selected into two groups receiving up to 30 sessions of either daily ( n = 101) or twice daily ( n = 109) 10 Hz rTMS to the left dorsolateral prefrontal cortex (DLPFC). The a priori primary outcome measure was remission rate as measured by pre and post treatment HAMD-17 scores. Length of hospital stay was a secondary post hoc outcome adopted due to the importance to cost of acute psychiatric care. Results: Remission rates were similar across groups, with 44.9% and 45.4% for twice daily and daily rTMS groups respectively, although these may be confounded by patient expectations, other treatments and medication changes given the naturalistic setting. The length of hospital stay was 10.11 days and 18.44 days for twice daily and daily rTMS respectively – the twice daily rTMS length of hospital stay was 45.1% shorter 95% CI [38.7% - 51.56%]. Dropout rates were high; Twenty-seven (24.77%) twice daily participants dropped out before 20 sessions were completed, and 35 (34.65%) of daily participants. Conclusions: Twice daily 10 Hz left sided rTMS remission outcomes were similar to traditional once daily rTMS but required a shorter length of hospital stay. This finding has substantial cost of care implications. If these findings are independently replicated, twice daily rTMS may become the standard of care for inpatient rTMS.


Introduction
Depression is one of the leading causes of disability worldwide (Friedrich, 2017). In Australia, the lifetime prevalence of depression is 1 in 7 (Australian Bureau of Statistics, 2007). Aside from the very significant morbidity costs stemming from diminished productivity, there are significant direct care costs associated with treating depression, with AUD$11 billion spent on mental health-related services in Australia in 2019-2020, a 5.4% increase since 2015-2016 (Australian Institute of Health and Welfare, 2022). It is anticipated this burden of disease will be aggravated by the recent pandemic and its lingering psychosocial impacts (Hajek et al., 2022).
Despite nearly a century of research into pharmacological treatments for depression, around 30% of patients consistently fail to respond to adequate trials of psychotherapy and/or antidepressants (Gaynes et al., 2009;Zhdanava et al., 2021;Rush et al., 2006). Electroconvulsive therapy (ECT) remains the gold standard form of brain stimulation for severe major depression due to superior efficacy. ECT unfortunately remains highly stigmatised, requires repeated general anaesthesia, and may be complicated by frequent adverse effects (Andrade et al., 2016). As such, there exists a significant population suffering from moderate to severe depression, unsuitable for ECT but resistant to pharmacotherapy and non-pharmacological approaches alike. For this population rTMS is an attractive option, but access and costs remain barriers to wider deployment globally.
rTMS is a safe, non-invasive, and non-convulsive form of neurostimulation. Electromagnetic pulses generated via a coil placed on the scalp induce a reciprocal electrical impulse in the underlying neural tissue. This leads to changes in neuronal excitability and metabolism (Miron et al., 2021a;Nahas et al., 2001). Depending on the protocol, stimulation can cause enduring excitatory or inhibitory modulation of the brain, suggesting rTMS may play an important role in many conditions underpinned by over or underactive neural circuits (Miron et al., 2021a). rTMS is generally well tolerated with the main side effect being mild pain or discomfort at the scalp (Miron et al., 2021a;Fitzgerald and Daskalakis, 2022).
Optimising the parameters of rTMS treatment protocols (and the equipment itself) remains an area of active research. The most thoroughly evaluated rTMS protocol consists of short bursts of 10 Hz pulses delivered to the Left dorsolateral prefrontal cortex (DLPFC) over 20-30 min known as high frequency (HF) rTMS (Miron et al., 2021a). A course usually lasts for 4-6 weeks of daily sessions, given in the inpatient or outpatient setting (McClintock et al., 2017). Evidence for Low Frequency (LF; 1 Hz targeting the right DLPFC) protocols suggest similar efficacy to HF rTMS but with advantages of shorter treatment sessions and simpler, less expensive coils (Miron et al., 2020;Miron et al., 2021b;McClintock et al., 2018;Modak and Fitzgerald, 2021). Likewise, intermittent theta burst stimulation has demonstrated non-inferiority to HF but with shorter treatment sessions (Blumberger et al., 2018). Increasing the number of rTMS sessions per day may have potential to increase accessibility of rTMS by compressing total course duration. While there are several meta-analyses of accelerated rTMS (2 or more treatment sessions per day) compared to sham for depression, currently there are limited studies comparing once daily to 2 or more sessions daily of rTMS (Sonmez et al., 2019;Fitzgerald et al., 2018;Theleritis et al., 2017;Modirrousta et al., 2018;Fitzgerald et al., 2020). The aim of this retrospective cohort analysis is to compare once daily to twice daily rTMS in a naturalistic setting to help further inform clinical utility and health economic benefit estimates.

Ethics
Ethics approval was obtained for a retrospective analysis of quality assurance data through a National Health and Medical Research Council (Australia) registered human research ethics committee (EC00380, project 323). All participants provided written, informed consent for the rTMS treatment, and consented to their de-identified data being recorded and used for quality assurance analysis. Ongoing quality assurance data analysis is a requirement for rTMS clinical services in Australia.

Participants and procedures
Participants consisted of a naturalistic cohort of patients admitted voluntarily to a private psychiatric hospital, with a diagnosis of Major Depressive disorder (MDD), between 2008 and 2018. A total of 209 participants had their de-identified data analysed for this study. Most participants were female (162 female, 47 male) with an average age of 46.4 years ( Table 1).
Collated data for analysis consisted of patient demographics, date of admission, insurance type, number of seizure-threshold raising drugs (defined as any benzodiazepine or anticonvulsant administered within 24 h of the treatment), if the participants had previous rTMS, whether they were allocated to once or twice daily groups. Participants had generally been referred by their GP or private psychiatrist and on admission had a thorough psychiatric evaluation to ensure clinical suitability for rTMS. Participants elected either once or twice daily treatments in discussion with the treating psychiatrist based on clinical equipoise and with awareness of the differing course duration and depth of evidence (greatest for once daily). Given the naturalistic setting, patients generally received concurrent psychiatric care including psychotherapy, medication changes, social worker input, etc. Generally, during the rTMS treatment no other meaningful changes to treatments were made. The treating psychiatrist of each patient referred to the rTMS services was advised by the rTMS psychiatrist not to make any substantive medication changes the week prior to rTMS commencing, and if an antidepressant medication or augmenting agent had been adjusted, at least three weeks had elapsed to gauge efficacy before the patient was considered for rTMS. In this way, any benefits from rTMS likely reflect either the therapeutic efficacy of the rTMS or spontaneous improvements, or both.
The patients were administered the rTMS treatment sessions by a rTMS credentialled registered nurse, with oversight by an experienced rTMS credentialed psychiatrist who conducted the titration to determine resting motor threshold and coordinates of the DLPFC. The rTMS stimulators were manufactured and calibrated by Medtronic. The HF rTMS protocol delivered pulses at 10 Hz frequency, with an amplitude of 120% of motor threshold. The resting motor threshold was titrated via visual observation of thenar eminence twitch, according to established protocols (Pridmore et al., 1998). 125 trains of 45 pulses each were delivered with 15 s intertrain interval for a total stimulation duration of just under 41 min over an approximately 60-min total session duration. The pulses were delivered to the Left DLPFC which was located via the Beam F3 method (Beam et al., 2009). For the twice daily group, treatment sessions were delivered with an interval of between 2 and 6 h depending on stimulator availability. The number of treatment sessions each participant received was variable and at the discretion of the treating psychiatrist based on overall clinical care needs of the patient as per naturalistic care (Table 1). Participants could discontinue at any time but were encouraged to have at least 15 sessions. Note that due to the naturalistic setting there was no systematic screening for the presence of adverse events nor reason for patient ceasing treatment early.

Instruments
The HAMD-17 rating scale was administered at baseline (just prior treatment commencing) and post completion of the course of treatment in all patients receiving care by the rTMS service at the clinic. The interval between baseline and endpoint measurement is recorded as the length of stay in days (LOS; see Table 1).

Inclusion / exclusion criteria
Only those participants admitted for their initial course of rTMS, and only patients whose HAMD-17 rating scales were administered by a clinician trained in their administration, with documented pre-and posttreatment scores available were included in the analysis. Those failing a standard rTMS safety checklist were excluded as per usual practice. There were no general medical (e.g., chronic pain) or psychiatric comorbidity (e.g., personality disorder) exclusions.

Outcome measures
The primary outcome measure was treatment remission (defined as a reduction in HAMD-17 score to below or equal to 7). Treatment response was defined as a 50% reduction in HAMD-17 scores. Inter-rater reliability assessment was not conducted. The secondary outcome measures were rate of drop out prior to course completion (20 rTMS stimulations) and total length of hospital stay. Participants were admitted prior to commencement of rTMS but of variable duration of stay, reflecting the naturalistic setting where sometimes patients elect to discharge early or need to remain after the end or rTMS if it was ineffective and subsequent inpatient care was needed.

Statistical analyses
We performed statistical analyses examining the non-inferiority of single-dose rTMS according to three hypotheses; namely, that twice daily rTMS is superior to traditional daily rTMS for the reduction and remission of depressive rates as measured by the HAMD-17, and that twice daily rTMS is superior for reducing HAMD-17 scores over time compared to traditional daily rTMS. We tested the first two hypotheses using chi-square for differences in proportions and calculated the relative rate ratios (RR) for the likelihood of a participant being classified as demonstrating remittent or responsive HAMD-17 scores according to the employed rTMS protocol. The third hypothesis was tested using ANCOVA to examine for between-groups endpoint differences in HAMD-17 after adjusting for baseline values, and rates of within-group change were examined using repeated-measures ANOVA. Repeatedmeasures effect sizes were calculated as Cohen's d, adjusted for the pre-post correlation, and interpreted using Ferguson's guidelines (Ferguson, 2009). Ferguson states that small, moderate, and large Cohen's d values are approximately 0.41, 1.15, and 2.7. Univariate normality for skew and kurtosis for the HAMD-17 were evaluated at both timepoints through standardised scores, and homogeneity of error variance was evaluated using Levene's test. No significant deviations from the assumptions of ANOVA were observed. We examined the length of stay between treatment conditions using the Mann-Whitney U test and evaluated the effect size using rank biserial correlation.

Comparison of treatment conditions for change in HAMD symptoms
To test the hypothesis that twice daily rTMS was superior to daily rTMS for the reduction of HAMD symptoms at the endpoint of the study, we performed a between-groups ANCOVA, adjusted for baseline HAMD symptoms. We observed no evidence of between-groups difference at the study endpoint: F (1, 207) = 0.508, p = .476, η 2 < 0.01. See Fig. 2 below.
Finally, we performed a repeated-measures ANOVA to test the hypothesis that twice daily rTMS was superior to daily for the reduction of HAMD symptoms over time. We observed a substantial main effect for time, and both groups reported large reductions in their HAMD symp-

Discussion
This study represents the largest naturalistic dataset to date comparing traditional (daily) vs twice daily rTMS. Despite important methodological limitations in terms of the non-experimental design, naturalistic quality assurance cohort data, and single site recruitment, the response and remission rates are consistent with other studies including those with more rigorous experimental designs (Sonmez et al., 2019;Dowling et al., 2020). While the daily and twice daily treatment effect sizes are large, they are consistent with the comparable literature, including two Australian naturalistic observational studies in a similar  private hospital setting (Dowling et al., 2020;Pridmore et al., 2020). Additionally, while not statistically significant the number of dropouts in the twice daily group were lower numerically than the daily group. Although the reasons for participants dropping out were not systematically captured, this plausibly suggests shorter admission durations are more tolerable to patients, as expected. The dataset herein provides an important empirical basis upon which to estimate health economic benefits of recommending twice daily rTMS for inpatients requiring such care. The improved affordability may enhance access to rTMS in settings where it was previously deemed too expensivethe main stumbling block for wider availability of rTMS (Cabrera et al., 2022;Pridmore and Pridmore, 2021).
Our findings suggesting non-inferiority of the daily and twice daily protocols is plausible considering our findings in the context of other recently published relevant literature (Sonmez et al., 2019;Fitzgerald et al., 2018;Theleritis et al., 2017;Modirrousta et al., 2018;Fitzgerald et al., 2020). While the older protocol used here is still in wide use, there are newer protocols under investigation including neuronavigated and intermittent theta burst-rTMS with preliminary evidence suggesting a significant increase in both efficacy and time to remission over the protocol used in this study (Modak and Fitzgerald, 2021;Cole et al., 2020). While these neuronavigated rTMS findings are exciting, such protocols require costly functional MRI facilities. Further refining the established protocols for rTMS offers an opportunity to improve the health economics of rTMS whilst not straying very far from the depth of existing research literature supportive of rTMS in the treatment of depression not responding to initial care steps.
Our analysis has several important limitations. This was a single site, open label, non-randomised non-sham-controlled 'quality assurance' cohort. It was subject to regression to the mean on HAMD-17 rating scales, confounding by milieu effects of hospital care and placebo effects with no sham comparator limb to discern such (Burke et al., 2019). Without randomisation there is a possibility of selection bias. Furthermore, rTMS was deployed as an add-on to treatment as per naturalistic clinical practice, meaning other therapeutics such as medications (including differences in antidepressant prescribing between groups) and counselling could have played a part in the outcomes. However, significant depression pharmacotherapy changes were discouraged prior (at least 3 weeks before rTMS) to all patients being referred for rTMS. Ward milieu and psychosocial group programs were provided equally to all patients as part of their care. Unfortunately, the proportion of the cohort with unsuccessful trials of antidepressants, nor the duration of the current depressive episodes were captured, when it could be expected these would influence the primary outcome. Of note, newer treatments such as esketamine were not available as treatment options over the study period, nor had rTMS gained approval for billing through Medicare. While changes to care during rTMS were discouraged, care was left entirely to the treating psychiatrist and thus other treatment adjustments mediating outcomes cannot be fully excluded.
Being quality assurance data obtained from an inpatient setting, participants were free to discharge from hospital or elect to discontinue, and a significant proportion, between 1 in 4 and 1 in 3 participants ceased care before a full course of 20 sessions. It is plausible that patient preferences and expectations have biased the effect of the twice daily group. For example, a proportion of participants may have selected the twice daily group due to their preference of a shorter admission and experienced rapid gains in line with their expectations. Unfortunately, data summarising the reasons for this high dropout rate were not captured.
Additionally, the lack of a placebo control may have introduced bias (Dechartres et al., 2013). This study also did not address inter-rater reliability, nor systematically screen for adverse events, nor relate these to subjects who ceased prior to a full treatment course. As such, no inference can be made about the safety or tolerability of once vs twice daily treatments. Of note, the lack of data on the time course of the response and the lack of follow-up beyond routine clinical care, leave open the possibility that the twice daily protocol, while acting quickly may also lead to rapid relapse. The Durability of rTMS effect is an important consideration on the need to progress to maintenance rTMS and the attendant costs (Senova et al., 2019a). Several studies have demonstrated that the antidepressant effect of rTMS endures for at least 6 months (Arici et al., 2022;Senova et al., 2019b). however, evidence regarding maintenance protocols for rTMS remains scarce (d'Andrea et al., 2023).
Notwithstanding these various important limitations, the naturalistic nature of the relatively large dataset herein is a strength. More naturalistic datasets are better generalisable to real world settingsand as such are often preferred by health economic modellers (Johnston et al., 2019). We believe this study helps further inform the field and adds to existing naturalistic cohort study data sets to assist both health economic modelling, and clinical practice.
With pressure on healthcare budgets necessitating optimal cost efficacy, reducing length of stay in hospital is an important issue. Halving the treatment time to complete a course of twenty rTMS sessions could halve the cost of care when this modality is delivered in an inpatient setting. Private inpatient psychiatric care costs up to AUD$700 per night, and so our observed reduction in LOS from an average of eighteen days to ten could save around $5600 in costs (RANZCP., 2017). Although many clinics performing rTMS in Australia do so with an outpatient model, this still involves a considerable travel burden to the patient, and this model is limited to those well enough for ambulatory care. Furthermore, the cost-benefit analysis for rTMS is likely to be more favourable for the more severe cases of depression where inpatient containment is required, meaning the above figure may be conservative. Thus, the present dataset may help inform policy and practice on more cost-effective ways to deliver rTMS to inpatients. The findings of this study, when taken in context of the ongoing advances in rTMS protocols, suggest there is scope to improve the health economic utility of rTMS through intensive care protocols.

Ethics
Ethics approval was obtained for a retrospective analysis of quality assurance data through a National Health and Medical Research Council (Australia) registered human research ethics committee (EC00380, project 323). All participants provided written, informed consent for the rTMS treatment, and consented to their de-identified data being recorded and used for quality assurance analysis. Ongoing quality assurance data analysis is a requirement for rTMS clinical services in Australia.

Declaration of Competing Interest
RB has no conflicts of interest to declare. DRS is a current investigator with the NHMRC Medical Research Future Fund (APP1200214), and has no conflicts of interest to declare.