Modest therapeutic effects of low-frequency transcutaneous electric nerve stimulation on insomnia among older adults: A 4-week multi-center, randomized controlled study

A 4-week, multi-center, randomized controlled study was conducted to evaluate the therapeutic ecacy and safety of low-frequency transcutaneous electric nerve stimulation (LF-TENS) for insomnia disorder. A total of 160 individuals aged 40 to 80 years with insomnia disorder were included and randomized to the experimental group receiving active device (n = 81) or control group receiving sham device (n = 79). Both groups used the device for four weeks, more than ve days a week. The participants also completed pre-and post-intervention assessment with questionnaires, sleep diaries, wrist actigraphy, and blood tests. We found a signicant improvement of subjective sleep quality, depressive/anxiety symptoms, and blood cortisol level in both experimental and control groups. Notably, among patients aged over 60 years, the experimental group showed better improvement after intervention in the change of PSQI score (2.60 ± 0.46 vs. 1.22 ± 0.46, p = 0.039; Cohen’s d = 0.99 vs 0.45) and blood cortisol level (-1.97 ± 0.53 µg/dl vs. -0.16 ± 0.53 µg/dl, p = 0.007; Cohen’s d = 0.56 vs 0.05). We found that LF-TENS was effective in improving subjective sleep quality in older adults aged over 60 years. The hypothalamic-pituitary-adrenal axis might be related to the therapeutic mechanism. to between-group differences after controlling for age, and baseline


Introduction
Insomnia disorder is the most common sleep disorder characterized by dissatisfaction with sleep quality or quantity that causes considerable daytime functional impairment. Its prevalence has been estimated ranging from 10 to 33 percent in adults according to the reports of various studies conducted worldwide [1][2][3][4] . Moreover, a longitudinal study revealed that 46% of insomniacs complained of persistent insomnia symptoms even after three years, suggestive of chronicity of the disease 5 . Insomnia disorder is also associated with physical and mental health problems such as cardiovascular diseases 6 , depressive disorder 7 , and cognitive impairment 8 .
The currently established treatment for insomnia disorder is restricted to short-term pharmacotherapy and cognitive behavior therapy (CBT), along with modi cation of precipitating and perpetuating factors for insomnia 9 . However, these conventional therapies have some limitations. The most widely used medications for insomnia including benzodiazepine receptor agonists have potential risk of side effects, such as dependency, tolerance 10 and cognitive decline 11 . CBT is costly and time-consuming because regular in-person encounters with therapists and good adherence of patients to the treatment sessions are required. Furthermore, over 40% of patients could not achieve remission with combined pharmacotherapy and CBT for six months 12 . To overcome these shortcomings of conventional therapies and to achieve better treatment outcomes, clinicians have tried to nd an alternative therapeutic approach for insomnia.
Cranial electrotherapy stimulation (CES) is an electrical neurostimulation technique that delivers a pulsed, alternating low-intensity electric current across the head via electrodes on the earlobes or scalp. It was rst introduced as a novel therapeutic modality for insomnia in the 1970s and since then, a couple of studies have indicated that CES could alleviate insomnia symptoms and improve sleep quality [13][14][15] . Meanwhile, some other researches have failed to con rm the therapeutic effects of CES on insomnia 16,17 and a recent systematic review also revealed inconclusive results regarding the e cacy of CES for insomnia treatment 18 .
Transcutaneous electric nerve stimulation (TENS) is a different type of neurostimulation modality which applies electric current pulses by electrodes through peripheral skin surface other than the head. TENS was found to be effective for pain control in various pain conditions 19 and it is widely used for relieving neuromuscular pain with only minor adverse effects, including mild erythema and itching sensations underneath the electrodes 20 . In particular, low-frequency (< 1000 Hz) TENS (LF-TENS) is effective in local stimulation of the application site without cardiovascular side effects compared to high-frequency TENS (HF-TENS) 21 . Considering that chronic insomnia disorder is associated with inappropriate physiological hyperarousal 22 and that the possible analgesic mechanisms of LF-TENS include the activation of gammaaminobutyric acid (GABA) pathway 19 , LF-TENS could be helpful for treating sleep disturbances. However, to our best knowledge, only few studies has examined the effects of LF-TENS on sleep 23,24 and no randomized controlled trial on this issue has been conducted.
To investigate safety and therapeutic effects of LF-TENS on insomnia, we have performed an open trial with 40 patients suffering from chronic insomnia. Our data demonstrated that four weeks intervention of LF-TENS can improve subjective sleep quality without intolerable or persistent adverse reactions 24 . However, the methodological shortcomings of this pilot study such as the absence of control group and objective sleep measurements may limit the interpretation of results. Hence, in the present study, we aimed to verify the safety and e cacy of LF-TENS application for the management of insomnia disorder, using a randomized double-blind and placebo-controlled parallel study design.

Baseline demographic and clinical characteristics
A total of 196 patients suffering from insomnia disorder were enrolled in the study and 176 patients were randomized. Of those, 90 and 86 persons had been assigned to the experimental group and the control group, respectively. The dropout rate was 10.0% (9/90) in the experimental group and 8.1% (7/86) in the control group, with no signi cant difference between the two groups. Finally, 160 patients completed all the study protocols and were included for the analysis (Fig. 1). When we compared the baseline demographic and clinical characteristics including objective sleep data from the PSG records, no statistically signi cant difference was found between the experimental and control group except for BMI and periodic limb movement index (PLMI). Compared to the control group, the experimental group showed signi cantly lower BMI and higher PLMI at the baseline assessment (Table 1). In addition, we found no signi cant difference in all baseline characteristics between those participants completed the study protocol and those who dropped out (data not shown).

Safety reports
Tolerability data were available for 175 participants because one participant had dropped out before the rst assessment for adverse events. Adverse reactions were reported by 35 patients (20%) at least once during the treatment period; headache (n = 7), chest discomfort (n = 2), myalgia (n = 5) skin rash (n = 3), urticaria (n = 1), itching (n = 9), tingling sensation (n = 6) dizziness (n = 3), epigastric pain (n = 2), nausea (n = 1) and fatigue (n = 3). The symptoms obviously not related to the device use, such as common cold, gastrointestinal disturbances due to food intake, and pain induced by other underlying medical conditions, were not regarded as adverse reactions. Three patients had dropped out due to adverse events; two in the experimental group (tingling sensation and back pain) and one in the control group (shoulder pain). All the adverse reactions were tolerable and self-limiting within several days for remaining participants. Also, no signi cant difference was found in the occurrence of adverse events between completers and drop-outs

Effects of LF-TENS intervention
There was no substantial difference in the number of device use (24.05 ± 3.19 days vs. 23.94 ± 2.66 days, p = 0.809) and the duration of daily use (53.74 ± 6.12 minutes vs. 54.50 ± 5.88 minutes, p = 0.422) between the experimental group and the control group. The average current intensity of LF-TENS was 1.12 ± 0.13 mA at the neck and 1.13 ± 0.13 mA at shoulder. Regarding rescue medication administration, those who used the active device showed slightly lower proportion of zolpidem use [35.9% (28/81) vs. 43.4% (34/79), p = 0.272] and lower dosage of zolpidem (9.51 ± 16.82 mg vs. 14.37 ± 20.28 mg, p = 0.100), although statistically not signi cant.
We tested the changes in the outcome measures before and after intervention period in the experimental and control groups among the whole study population. After four weeks of intervention, signi cant reductions of the PSQI, BDI, and BAI score were found in both study arms. Although the amplitude of these improvements was greater in the experimental group, it did not reach statistical signi cance ( Table 3). The positive treatment response rate was 37.0% (30/81) in the experimental group and 30.4% (24/79) in the control group, yielding no substantial difference between the two groups (p = 0.373) ( Fig. 2A). As for sleep parameters, both the experimental and control groups have reported an evident improvement in all the subjective sleep quality based on sleep diaries, without distinct between-group difference. In blood test results, meaningful differences before and after treatment were observed in serum cortisol levels both in the experimental and control groups, producing no signi cant difference in the changes of serum cortisol levels between the two groups. There was no profound change in actigraphy after study protocol in both study groups, without any noticeable between-group difference. Data are presented as mean ± standard deviation. † Paired t-test was performed to examine within-group differences ‡ Analysis of covariance was performed to examine between-group differences after controlling for age, sex, and baseline BMI.
PSQI, Pittsburgh sleep quality index; ESS, Epworth sleepiness scale; BDI, Beck's depression inventory; BAI, Beck's anxiety inventory; NRS, numeric rating scale; SL, sleep latency; WASO, wake after sleep onset; TST, total sleep time; SE, sleep e ciency; IL-6, interleukin-6; CRP, C-reactive protein; TNF-alpha, tumor necrosis factor-alpha.   (Fig. 2B). Figure 3 represents the changes in PSQI score after intervention for each participant. Regarding blood test results, serum cortisol levels were signi cantly decreased in the experimental group (10.51 ± 2.90 µg/dl to 8.87 ± 2.94 µg/dl, p = 0.004) but not in the control group (10.58 ± 2.66 µg/dl to 10.42 ± 3.52 µg/dl, p = 0.775), which also resulted in a substantial difference in the decrease of cortisol concentration between the two group (p = 0.007). Although statistically insigni cant, we found a trend-level decrease of serum TNF-alpha levels in the experimental group (0.92 ± 0.32 pg/ml to 0.85 ± 0.31 pg/ml, p = 0.073), while an increase was observed in the control group (0.82 ± 0.28 pg/ml to 0.88 ± 0.39 pg/ml, p = 0.131). Also, a signi cant inter-group difference was found between the two groups in the change of serum TNF-alpha concentration (p = 0.007). Compared to the control group, the experimental group showed larger estimated effect size in the serum cortisol (Cohen's d = 0.56 vs 0.05) and TNF-alpha levels (Cohen's d = 0.22 vs − 0.18).
Among the middle-aged patients aged 40 to 60 years, we could not nd a statistically meaningful difference in the changes of outcome measures between the experimental and control groups (Table 4). In addition, no evident difference was observed both in the positive treatment response rate across the two groups [31.7% (13/41) vs. 39.5% (15/38), p = 0.471] (Fig. 2C).

Discussion
This study demonstrated the safety and therapeutic effect of low-frequency electrical stimulation for elderly insomnia patients without comorbid medical or psychiatric conditions or certain sleep disorders. We found a greater improvement in subjective sleep quality and a higher treatment response rate in those with LF-TENS treatment compared to those with sham treatment, in older insomnia patients over 60 years of age. In line with this nding, among the elderly patients, a signi cant decrease in serum cortisol level was observed only in the experimental group. With regard to safety issues, no serious adverse reactions had occurred during the study period and the risk of adverse events of LF-TENS treatment did not differ from that of sham treatment.
Our ndings demonstrated the e cacy and safety of electrical stimulation therapy targeting on the peripheral muscle, without direct stimulating effect on central nervous system (CNS) activity. In this study, bilateral upper trapezius muscles were selected as application sites for several reasons. Since the trapezius muscles are innervated by spinal accessory nerve and cervical nerves C2 to C4 36 , the therapeutic effects of LF-TENS on trapezius muscles may affect CNS via this neuronal pathway. This muscle is also known to be related to sleep disturbances. A prospective study reported that an increased muscle response in the upper trapezius muscle could be a strong predictor of sleep complaints 37 . In addition, chronic trapezius myalgia might be associated with anxiety and depression 38 , which may contribute to the aggravation of insomnia symptoms. Moreover, in consistent with our ndings, our preliminary study revealed the therapeutic effects of LF-TENS administration on the trapezius muscles without crucial adverse reactions 24 .
The exact mechanism behind the effects of LF-TENS on insomnia still remains unclear yet. One possible explanation could be that well-established analgesic effects of LF-TENS 19,20 may help ameliorate insomnia symptoms. Chronic pain is thought to be closely related with sleep disturbances via hyperarousal following the experience of pain 22 and proper management of chronic pain is important for the management of insomnia symptoms 39,40 . In the current study, however, no signi cant post-intervention change of pain intensity was observed for both between experimental and control groups in the elderly as well as in the middle-aged. Additional analysis yielded the similar results after exclusion of patients with adverse events that can affect the pain intensity (data not shown). Our pilot study also revealed no signi cant difference in pain severity before and after LF-TENS 24 . These ndings may, to some extent, result from the low baseline pain severity in our study participants since we have excluded those with medical or psychiatric comorbidities that can be accompanied by severe pain. Our results also suggest that the therapeutic e cacy of LF-TENS could be related to other physiological mechanisms.
Another hypothesis would be that the hypothalamic-pituitary-adrenal (HPA) axis activity might be connected to the working mechanism of LF-TENS for insomnia. The existing literature has documented the bidirectional relationship between HPA axis hyperactivity and insomnia 41,42 , and LF-TENS can downregulate HPA axis activity since muscle relaxation that might be achieved with LF-TENS, was reported to reduce cortisol secretion 43 . It is well-known that cortisol acts as a marker of HPA axis function since it is released from adrenal gland in response to physical and psychological stress and its serum concentration is regulated by HPA activity 44 . On the basis of the aforementioned grounds, we hypothesized that serum cortisol levels would be decreased after LF-TENS treatment. As expected, we found a signi cantly more remarkable decrease in cortisol levels in the experimental group compared to the control group among the elderly patients, while no signi cant between-group difference was found in the middle-aged patients.
These results support evidence that the therapeutic e cacy of LF-TENS may arise from the modulation of HPA axis activity, since this hypothesis can explain the reason why LF-TENS was more effective among older patients in our study. Unfortunately, only few studies have dealt with the effects of TENS on HPA axis activity in humans and they have shown contradictory results. Chu and colleagues have reported increased salivary cortisol levels after TENS 45 , while a study of the effects of TENS on the plasma concentration of cortisol failed to nd a signi cant difference in cortisol levels 46 . For these reasons, further investigations are needed to clarify the exact physiological process underlying the clinical bene ts of LF-TENS on insomnia.
To the best of our knowledge, our pilot study was the only study assessing the e cacy of LF-TENS for insomnia patients with no underlying disease, representing the response rate of 57.5% in chronic insomnia patients aged 55 years or older 24 . In the current study, the positive treatment response rate was 37.0% for insomnia patients aged 40 to 80 years. This relatively low response rate may be attributable to the age difference in study participants and more strict de nition of positive treatment response in the current study. Moreover, we found that 42.5% of elderly insomnia patients aged over 60 years had achieved optimal treatment outcomes with LF-TENS, representing signi cantly better improvement of insomnia symptoms compared to those with sham treatment.
We observed the therapeutic e cacy of LF-TENS among participants over 60 years of age while no The strengths of this study include the large study population consisting of insomnia patients without medical and psychiatric comorbidities which may affect the sleep quality. Moreover, our study is the rst randomized-controlled trial providing some evidence for the use of LF-TENS for the treatment of insomnia among the elderly. Yet, the current study has several limitations to be considered. First, we cannot generalize our ndings to severe insomnia patients requiring ongoing treatment since we excluded those with recent insomnia treatment history. However, co-existing insomnia therapies may blur the effect of LF-TENS and thereby weaken the clinical implications of the study. Second, all the study participants were 40 years or older, and thus our results cannot be generalized for all ages. Finally, there were individual differences in adherence to treatment, which might affect the outcome of our analysis. But these differences could not be well controlled, and also there was no signi cant difference in compliance between the experimental and control group.
In conclusion, LF-TENS was safe and showed modest effects for the treatment of insomnia in elderly patients over 60 years of age. The modulation of HPA axis activity after intervention might be related to the therapeutic e cacy. Physicians may consider LF-TENS as a novel and safe therapeutic option for the management of insomnia in the elderly insomnia patients. A larger scale, highly controlled human studies are required to provide better insight into the therapeutic e cacy and long-term consequences of LF-TENS for the management of insomnia disorder.

Study population
The current study was conducted at two medical centers, Seoul National University Bundang Hospital and

LF-TENS Intervention
As a LF-TENS device, we used CR-9 ® medical device manufactured by Crown Medical, Inc. (Seoul, Korea). It was approved by the Korea Ministry of Food and Drug Safety and is currently commercially available for relieving muscular pain. The stimulator delivers an alternating current of < 1 mA (peak output voltage ranges from 0.3 to 0.64 V) at a frequency of 400 Hz, with a pulse duration of 100 microseconds. Electric current is transmitted through three transcutaneous nickel plate electrodes attached to the back and neck area to stimulate both trapezius muscles. The sham device was created to look identical to the CR-9 device, but it has no electrical stimulation function.
Eligible participants were randomly assigned in a 1:1 ratio to either the experimental group receiving an active device, or control group receiving a sham device. The random assignment was performed via the strati ed permuted block randomization method, using SAS version 9.4 and both patients and investigators were blind to the intervention condition. After assignment, patients were instructed to wear the device for an hour just before bedtime, more than ve days weekly, for four weeks. Although the optimal dose and duration of LF-TENS application for insomnia treatment were not established, our preliminary study demonstrated the modest e cacy and safety of CR-9 ® medical device for improving chronic insomnia, using the same protocol of device use 24 . A recent review has also indicated that su cient therapeutic effects could be attained by daily use of electrical stimulation for more than four weeks, with a session duration of 20 to 60 minutes 27 . The patients were allowed to take a rescue medication (zolpidem 5 to 10mg) if they experience the aggravation of insomnia symptoms for more than ve consecutive days or reduction of total sleep time below three hours at least one day. For verifying the safety of the device use, we have checked the occurrence of any adverse event after two and four weeks of intervention. Before and after treatment, participants were asked to complete self-report questionnaires concerning sleep quality, psychiatric symptoms, and pain. The pre-and post-intervention evaluations were performed in the morning after overnight PSG and in the last visit after the end of intervention, respectively. As for the primary outcome measure, the Pittsburgh Sleep Quality Index (PSQI) was adopted to assess subjective sleep quality 29 . Since there is no established de nition for clinically signi cant treatment response in insomnia 30 , we de ned the positive treatment response as a decrease in PSQI score ≥ 3 points after intervention, based on several preceding insomnia treatment studies 30,31 . Different from these studies, however, we did not adopt the criteria based on sleep diary data due to the unavailability of data for some participants. The secondary outcome measures were the changes in the scores of Epworth Sleepiness Scale (ESS) 32 , Beck Depression Inventory (BDI) 33 , Beck Anxiety Inventory (BAI) 34 , and Numerical Rating Scale for pain (NRS) 35 .

Sleep diaries and Actigraphy
Subjective and objective sleep parameters were assessed as secondary outcome measures. To obtain subjective sleep parameters, participants were asked to keep a sleep diary every day in the morning during the entire intervention period. The sleep diary contained items evaluating subjective sleep factors such as time in bed (TIB, the number of minutes spent in bed), sleep latency (SL, the number of minutes taken to fall asleep), wake after sleep onset (WASO, the number of minutes of wakefulness after sleep onset), TST, and sleep e ciency (SE, the ratio of TST to TIB). For examining the changes in subjective sleep quality before and after the treatment, we compared sleep parameters from sleep diaries during the pre-and postintervention actigraphic recordings. We excluded the sleep diary data from participants whose answers were obviously unreliable, for example, "TST = 0." To con rm the adherence to the intervention protocols, the duration of daily device use was also recorded in daily sleep diaries.
For the assessment of objective sleep parameters, our study participants were requested to wear an accelerometer (wGT3X-BT, ActiGraph, LLC, Pensacola, FL, USA) on their non-dominant wrist. Actigraphic monitoring was performed twice, before and after treatment. The pre-and post-intervention assessments were carried out for four days just before the start of intervention and for the last four days of the intervention period, respectively. The wrist actigraphy estimated sleep-wake status by capturing and recording physical activity and information regarding TIB was obtained from the daily sleep diary. ActiLife 6 software (ActiGraph, LLC, Pensacola, FL, USA) was used to analyze raw data and calculate the following four sleep variables: SL, WASO, TST, and SE. Data of participants who made proper use of actigraphy as instructed for at least 3 days for both times were included for the analysis.

Blood tests
The pre-and post-intervention blood tests were conducted as secondary outcome measures, concurrently with the questionnaire assessments. Blood samples were drawn from the antecubital vein in the morning (between 8AM and 9AM), in the fasting state more than 8 hours. They were properly processed and transported to the testing institute (Seoul Clinical Laboratories, Seoul, South Korea). We estimated the serum levels of Interleukin 6 (IL-6), Tumor necrosis factor alpha (TNF-alpha), glucose, C-reactive protein (CRP), cortisol, and insulin.

Statistical analysis
We compared demographic and clinical characteristics between the experimental and control groups, using the independent t-test for continuous variables and the chi-square test for categorical variables. Paired ttest or Wilcoxon signed rank test was adopted to examine the intra-group differences between pre-and postintervention assessments including questionnaires, sleep diaries, wrist actigraphy, and blood tests. To evaluate the inter-group differences in the changes of parameters after treatment between the two study arms, we used analysis of covariance (ANCOVA) adjusted by age, sex, and baseline BMI. The exploratory analysis was performed after the patients were divided into two groups: those aged > 60 years and those aged ≤ 60 years. Cohen's d was calculated to estimate the effect size of the intervention in each group. All statistical analyses were carried out using SPSS version 25.0 for Windows (IBM Corp., Armonk, NY, USA) and a two-tailed p-value of less than 0.05 was considered statistically signi cant.

Ethics statement
All the participants were informed of the purpose and procedures of the study and they provided written consent at the rst visit. All study procedures were conducted in accordance with the Declaration of Flowchart of study participants The positive treatment response rate in the experimental and control group among overall participants, elderly patients (aged > 60 years), and middle-aged patients (aged 40-60 years)