Cereset Research Standard Operating Procedures for Insomnia: A Randomized, Controlled Clinical Trial

This single site, randomized, controlled trial was conducted in the Neurology Department at Atrium Health Wake Forest Baptist, an academic medical center in Winston-Salem, North Carolina. Participants had symptoms of insomnia persisting for at least 1 month based on Insomnia Severity Index ≥8, and this was not attributable to another known diagnosis (such as exclusions listed below that could cause sleep disturbance). Participants were recruited by community physician referral, word of mouth, and advertisement. Potential participants were excluded if they were unable, unwilling, or incompetent to provide informed consent, or physically unable to attend study visits. Other exclusions were: known history of obstructive sleep apnea, periodic limb movements, seizures, urinary problems such as benign prostatic hypertrophy, severe hearing impairment, or ongoing need for treatment with opiates, benzodiazepines, antipsychotics, stimulants, thyroid hormone, antidepressants such as selective serotonin reuptake inhibitors (SSRI), serotonin-norepinephrine reuptake inhibitors (SNRI) or tricyclics, or sleep medications such as zolpidem or eszopiclone. Those with anticipated, ongoing use of recreational drugs or alcohol, and those with weight over 285 pounds (chair limit for intervention and placebo administration) were excluded. Individuals participating in another active intervention study, and individuals with previous history of receiving HIRREM/Cereset Research intervention were also excluded. Participants were instructed to abstain from alcohol or recreational drugs during the intervention and until final data collection; whether participants adhered to this instruction was assessed by the study coordinator at each data collection visit. Participants were advised to suspend chiropractic, cranial-sacral therapy, and bio-energy work during the intervention, and for at least 3 weeks following, and to refrain from caffeine use after 1:00 pm. All participants were instructed to otherwise continue their ongoing care (any other medications or therapies, outside of those listed above as exclusions).

This was a randomized, blinded, controlled trial. Study participants and all study personnel, except Technologists administering the intervention, remained blinded to group assignment. The protocol was approved by the Institutional Review Board at Wake Forest University Health Sciences, with full informed consent obtained from each participant. The 22 participants were randomly allocated 1:1 to intervention or a random tones control using blocked randomization with a block size of 4; these groups composed those analyzed in the intention-to-treat analyses. Standard intention-to-treat analysis is an approach in which all randomized participants are included in the statistical analyses even if they do not have follow-up data.²¹ The randomization scheme and assignments were created independently by a team member who had no contact with the participants and were securely maintained in a password-protected Excel file accessed only by the unblinded Technologist (at the time of the first intervention/control session following enrollment). Group assignments were made independent of and concealed from the team member enrolling the participant and conducting follow-up outcome assessments; thus, outcome collection was blinded to group assignment. The study had been approved to enroll up to 24 participants in order to achieve a target goal of 20 to complete the intervention and enrollment ended once there were 20 completers.

Twenty-two participants were randomized to receive the CR-SOP intervention. Ten were assigned to receive tones linked to brainwaves (LB, intervention), while 12 were assigned to receive random tones not linked to brainwaves (NL, control), in addition to continued current care. Written informed consent and all baseline measures were obtained during an enrollment visit (V1). Figure 1 shows an overview of the study schedule and activities.

Sessions began 0-14 days following the V1 enrollment visit. Participants received 10 intervention/control sessions over a 4-week period (S1-S10). Participants received the first 2 sessions on consecutive days. Afterwards, participants could receive 2 intervention/control sessions during a half-day period with an hour break between sessions. Participants received the third session within the first 5 days. Participants were encouraged to complete the remainder of the 10 intervention/control sessions with no more than 5 days between sessions and were otherwise scheduled based on convenience and scheduling needs for the participant. The time of day was noted for both intervention/control sessions and data collections. Participants completed an expectation measure where they were asked to guess study group assignment prior to Session 5.

Zero to 14 days after the final intervention/control session there was a post-intervention/control data collection visit (V2). All measures were repeated. Four to 6 weeks after completion of the intervention/control sessions, there was another post-intervention/control data collection visit (V3), which served as the primary outcome for the study and all measures were repeated, including the expectation measure regarding group assignment. The blind was then broken following data collection at the V3 visit, and group assignment was shared with the participant. Although official involvement in the study was completed after V3, those who were in the control group were offered a crossover course of CR-SOP sessions.

Participants who chose to crossover completed intervention sessions within 3 months of V3. The same scheduling timeline and data collection plan was followed for crossover sessions. A data collection visit (V4) occurred 0-14 days after intervention completion. A final data collection (V5) was conducted 4 to 6 weeks after the last intervention crossover session.

All participants received a standard intervention protocol series of 10 sessions of 60-75 minutes in length, consisting of 4-6 individual protocols, each typically lasting from 6-20 minutes. Total intervention listening time was 536 minutes; additional session time included time for placement changes and Technologist check-in with the participant.

During sessions, with the participant comfortably at rest, sitting or reclining, paired sensors were placed over specific target areas on the scalp corresponding with brain regions/lobes to be observed. The Cereset Research (CR) system includes the use of 64-bit processing architecture for faster feedback than the prior HIRREM technology, 4 sensors, and more standard protocols (while retaining flexibility with length and sequencing of the standard protocols). All protocols were done with eyes closed. Four sensors (2, paired locations) are placed on the scalp at a time. However, only 1 pair of sensors are actively echoing feedback. The software automatically switches from 1 pair to the other between protocols. This reduces the number of sensor placement changes needed during a session by half compared to the prior HIRREM version of the technology, resulting in shorter session time and fewer interruptions. The CR intervention is limited to 4 paired protocols during sessions targeted the bilateral hemispheres according to the 10-20 International System at F3/F4 and P3/P4, FP1/FP2 and T3/T4, C3/C4 and O1/O2, and AFZ/POZ and CB1/CB2.

For those in the intervention group, software algorithms identified dominant frequencies as specific frequencies in the 0.5 Hz to 48.5 Hz range and translated each to an audible tone in real time. These selected tones were chosen from a set of engineered tones whose ADSR (attack, decay, sustain, release) envelopes were all exactly equal in length, to eliminate 1 tone having more influence than another lower or higher tone. Each chosen tone was echoed back to the participant via ear buds with as little as 4-8 millisecond delay (Personal communication, Brain State Technologies). The volume at which each note is played is the same, and is based on the comfortable volume setting indicated by each participant, since CR is working to echo the brain rather than being any type of operant conditioning intervention.

Participants were thus able to “listen to their brain,” and figuratively speaking, to observe their brain pattern in an acoustic mirror, via tones linked in real time to the electroencephalographic pattern in the brain. CR-SOP software algorithms are intended to improve inflexible, potentially maladaptive patterns of cerebral activity. Particular attention is given to activity patterns suggesting dominant hemispheric asymmetries and/or suboptimal ratios of electrical amplitudes across the spectrum of frequencies.¹⁶ Pilot data suggest correlation of brain pattern with autonomic cardiovascular regulation outcomes (heart rate variability, HRV), and that changes in asymmetry of frequencies and amplitudes may be observed from pre- to post-Cereset Research intervention.^18,22

This closed-loop, recipient-unique, acoustic-stimulation brain feedback, or acoustic neuromodulation, supports the brain to auto-calibrate, self-adjust, relax, and to shift towards a more balanced state, often with reduced hyperarousal. This process is accomplished with no need for active, conscious, cognitive involvement by the participant, operant conditioning, or training the brain to accomplish anything.¹⁶

Those assigned to the control group received randomly generated tones during sessions, as a sham, active control intervention. All activities, procedures, and session times were similar to the intervention LB condition, with placement of sensors on various scalp locations. Sensors used for the control group had no active recording function and the tones were randomly generated, with no relationship to current brain activity.

All sessions (intervention or control) occurred with eyes closed, for which the participant was instructed to relax while sitting or reclining in a chair (Human Touch PC-6).

Study data were collected and managed using REDCap electronic data capture tools hosted at Wake Forest School of Medicine.²³ REDCap (Research Electronic Data Capture) is a secure, web-based software platform designed to support data capture for research studies, providing 1) an intuitive interface for validated data capture; 2) audit trails for tracking data manipulation and export procedures; 3) automated export procedures for seamless data downloads to common statistical packages; and 4) procedures for data integration and interoperability with external sources.

A series of measures were collected at the enrollment and follow-up visits, including self-report symptom questionnaires, and continuous recordings of BP and HR used to analyze measures of autonomic cardiovascular regulation. The primary outcome in this study was differential change in the score reported on the Insomnia Severity Index (ISI) from V1 to V3 (enrollment to 4-6 weeks post intervention/control completion, Figure 1).

Participants in the control group who chose to crossover to the linked to brainwave intervention also had 2 additional data collection visits as described above, V4 (0-2 weeks after active intervention completion) and V5 (4-6 weeks post-intervention completion). Other than the expectation measure, the measures collected at V4 and V5 were identical to V2 and V3.

The Center for Epidemiologic Studies Depression Scale (CES-D) is a 20-item survey assessing depressive symptoms to screen for depression.²⁷ Scores range from 0-60, with a score of 16 commonly used as a clinically relevant cut-off.²⁸

The Generalized Anxiety Disorder-7 (GAD-7) is a 7 item screening tool for anxiety that is widely used in primary care. GAD-7 is a brief, reliable and valid measure of assessing generalized anxiety disorder.²⁹ The GAD-7 score (the sum of 7 items scores, 0-3) can range from 0 to 21.

The Perceived Stress Scale (PSS) is a ten-item psychological instrument for measuring the perception of stress. It is a measure of the degree to which situations in one’s life are appraised as stressful. Items were designed to evaluate how unpredictable, uncontrollable, and overloaded respondents find their lives. The scale, with answers rated from 0-4, also includes a number of direct queries about current levels of experienced stress.²⁹

Participants were asked which group they believed they were assigned to, acoustic stimulation linked to brainwaves (LB, intervention), or non-specific acoustic stimulation (NL, control). This expectation measure regarding group assignment was obtained at V1, before the fifth session during the intervention period, and at the outcome visits. The second collection of the expectation measure, during the intervention period, but prior to anticipated meaningful benefit in the intervention group, allowed a realistic evaluation of the effectiveness of the blinding for the control intervention.

Continuous BP and HR were acquired from noninvasive finger BP cuff for a minimum of 10 minutes, with subjects lying down quietly, supine. Systolic BP (SBP) and beat to beat, RR intervals (RRI) files were generated via the data acquisition system (BIOPAC Acknowledge 4.2, Santa Barbara, CA) and analyzed using Nevrokard BRS software (Medistar, Ljubljana, Slovenia) for measures of BRS, HRV and BPV as previously published¹⁹ and described below.

Frequency Method. Power spectral densities of SBP and RRI were assessed over specified ranges (low frequency, LF: .04-.15 Hz; high frequency, HF: .15-.4 Hz). LF and HF alpha indices were calculated as measures of BRS. Power of RRI spectra in LF, HF range (LF_RRI and HF_RRI) were calculated as measures of HRV and the ratio of LF_RRI/HF_RRI is used as a measure of sympathovagal balance.³⁰

Time Domain Analysis was used to calculate BRS using sequence method as Sequence UP, DOWN and TOTAL. HRV was also measured as standard deviation of normal beat-to-beat interval (SDNN) and the root mean square of successive RRI (rMSSD). BPV was determined as the standard deviation of the mean arterial pressure (SDMAP).

Participants were asked about any new or worsened symptoms at every data collection visit, and during Technologist check-ins prior to intervention/control sessions. Safety monitoring was conducted by the Principal Investigator and adverse events were reported to the Institutional Review Board using established protocols.

Linear mixed models (LMMs) were used to contrast longitudinal changes in outcome measures between the intervention and control groups.³¹ LMMs provide a natural mechanism to address correlations induced by repeated measurements on a single participant as well as the potential presence of incomplete data due to participants that are lost to follow-up. The primary analytic model included fixed effects corresponding to group assignment, measurement time point, and their interaction. Mean contrasts were used to compare the change for the outcome measures between groups from baseline to the follow up assessments at V2 (0-2 weeks after randomized intervention/control completion) and V3 (our primary test of efficacy for the primary ISI measure; 4-6 weeks after completion of randomized intervention/control sessions). Among those randomized to sham control who chose to crossover to intervention, linear mixed models were used to estimate mean changes in measures from baseline to each follow-up visit and changes from V3 to the V4 and V5 post-crossover intervention visits. Following recent practical guidelines for LMMs,³² we used a combination of goodness-of-fit measures,³³ residual-based diagnostics, and outcome transformations to address important assumptions (homogeneity of the variance and normality for the model residuals) and specification of the covariance structure. The LMMs were fitted using PROC GLIMMIX in SAS. In addition to the primary intention-to-treat analysis, a per-protocol analysis was done as part of sensitivity analyses.

Of the 22 participants enrolled and randomized, 10 were assigned to the intervention and 12 to the control group; all were included in intention-to-treat analyses. Of those assigned to intervention, 9 completed all sessions (1 discontinued after 6 sessions due to schedule conflicts). In the control group, 11 completed all 10 sessions (1 discontinued after 1 session due to schedule conflicts). Overall 10 participants (7 in control group) reported minor violations to the alcohol abstinence guidance during the study period (3 or fewer drinks total among 7 participants, occasional consumption among others). Across the various data collection time points, 56% to 70% in the control group and 44% to 78% in the intervention group reported they felt they were receiving active intervention on the expectation measure, indicating adequate blinding.

Raw values of the sleep measures at baseline and follow-up are demonstrated in Table 2. The primary outcome for this study was differential change in ISI scores from V1 to V3. Among all 22 randomized patients (intention to treat linear mixed model analysis using the GLIMMIX procedure, Figure 3A), in the control group there was a mean reduction of ISI score of 4.69 points (SE 1.40, P = .003) and in the intervention group there was a larger and clinically significant reduction in ISI score (7.27 points, SE 1.55, P = .0002). The change in ISI score in the intervention group was not significantly different from the control group (P = .24 for difference between groups, 95% CI: (-7.05, 1.88) points for difference between groups). Change in excessive daytime sleepiness (ESS) did not differ between randomized groups. Sleep quality significantly improved in the intervention group compared to controls at V3, with adjusted mean change of −5.74 points on the PSQI instrument in the intervention group compared to control (P = .0007).

In the crossover group, once these individuals received the intervention, many of the scores improved, including the ISI, with mean reduction in ISI score of 4.33 points (SE 1.11) in the linear mixed model analysis from end of control follow-up (V3) to 4-6 weeks after completion of linked to brainwave intervention (V5) P = .013 (Figure 3B). Sleep quality improved, with mean reduction in PSQI from baseline (V1) to V4 of 3.17 points and 3.67 points to V5 (P = .03 and .02 respectively; there had been a non-significant change from baseline to V3 in this group during the sham control condition: 2.83 points, P = .11).

Table 2 also demonstrates raw mean baseline and follow up scores for anxiety, depression, and perceived stress. Some reductions in anxiety, depression and perceived stress scores were observed during follow up in both randomized groups. Specifically, in linear mixed model analysis, the intervention group had 50.8% reduction in GAD-7 anxiety score from baseline to V3 (P = .005) and 56.6% reduction in CES-D depression score from baseline to V3 (P = .003). Reductions in the control group from V1 to V3 were 33.2% and 26.2% for the GAD-7 and CES-D respectively. Perceived stress scale scores improved in both groups from V1 to V3, by 2.3 points in the intervention group and 3.9 points in the control group; the difference was not statistically significant. Among those who crossed over to receive active intervention, only the CES-D demonstrated a statistically significant improvement from V3 to V5 (P = .03).

Based on intention-to-treat analysis, significant interval improvements were observed across multiple measures of HRV (SDNN, and rMSSD) and BRS (Sequence ALL) at post intervention time points (V2, V3), compared to V1 in the intervention group (Figure 4A-C). The improvements in the intervention group were also significant when compared to outcomes in the NL control group at V2. In the control group, no significant changes were observed at V2 or V3 compared to V1 values. For those who participated in the crossover active intervention, measures of autonomic function then significantly improved at V5 compared to V3 (Figures 4D-F).

There was 1 dropout per arm, due to scheduling constraints/time required for the intervention, as mentioned above. There were no serious adverse events and no study related adverse events. One participant reported tinnitus, hot flashes, and increased pulse, but this was felt to be due to a change in medication/diet rather than the intervention.

The current study successfully administered CR-SOP sessions within a randomized, controlled trial design and demonstrated feasibility and clinically relevant magnitude of improvement in insomnia and sleep quality for this less operator-dependent innovation of the HIRREM intervention. The intervention was well-tolerated and the blind was maintained with the revised control condition. This study provides foundational preliminary evidence that scalable innovations of the technology are feasible and likely effective.

Our study has the following limitations. Participants with any level of insomnia were enrolled (ISI ≥8), rather than the more severe insomnia required for inclusion in our prior larger trial. While the magnitude of change in insomnia severity was similar in this small study compared to our prior larger study, and the confidence interval includes a clinically meaningful difference between groups, the confidence interval also includes smaller effects and no effect. Thus, reduced efficacy of the technique for milder insomnia has not been excluded. As this was a pilot study designed to demonstrate feasibility of the CR-SOP technique, and not a fully powered trial, it is not surprising that the primary outcome did not demonstrate a statistically significant difference between the intervention and control groups. However, the magnitude and direction of change in insomnia, sleep quality, and many of the autonomic measures were similar to the benefit demonstrated in our prior larger trial.¹⁹

Although heart rate variability and baroreflex sensitivity are objective outcome measures, this study did not collect direct measures of sleep such as polysomnography-based sleep stages and sleep spindles or slow-wave content, due to funding limitations. Within the growing non-pharmacological landscape for insomnia, there has been recent interest in using acoustic stimulation to enhance slow wave oscillations during sleep.⁴⁸ The oscillation literature suggests the pathophysiology of insomnia could be characterized by a lack of slow wave activity, thus measurement of slow wave sleep content would be beneficial in future studies.

No daily sleep diaries were required. Scores for depression and anxiety did not meet clinical criteria at baseline, so a floor effect may have affected those outcomes. Participants in this study had a mean age of >50 years, were mostly female, and of white, non-Hispanic ethnicity. This limits the generalizability of our findings, and future studies should include more diverse samples. More individuals in the intervention group were employed than in the control group; it is possible that job related impacts on sleep could have reduced the intervention’s potential benefit for sleep relative to those who were not actively employed. The use of several commonly prescribed categories of medications was also an exclusion, limiting generalizability. Other questions not addressed by this study include the implications of the present findings for individuals using psychotropic agents since these were excluded. Future studies should be conducted with expanded scope, including larger samples, with objective sleep measures included. These future studies could include longer follow up periods and potentially follow-up intervention sessions to evaluate whether this would extend the period of benefit.