This was a 12-week prospective, observational period following a clinical trial (NCT04916600) described elsewhere (36). Briefly, the clinical trial was a double-blind, randomized, stratified by sex, single-site trial at four facilities to compare whether active device stimulation led to greater reduction in withdrawal symptom severity during residential opioid discontinuation (31).

The trial was originally designed and powered to detect an effect of active versus sham device use on post-discharge substance use as the primary efficacy endpoint. However, following FDA discussion, it was decided to seek device clearance through the 510(k) pathway for opioid withdrawal reduction during the residential period (37). Although the clinical trial was limited to the first hour of the intervention, the post-discharge prospective, blinded monitoring was maintained.

For the clinical trial, all participants were screened, enrolled, and underwent evaluation at four (two male, two female) residential addiction treatment facilities within the same organization with typical length of stays of 28 days. Senior management, therapists, nursing staff, and core services were common across all locations. The admissions process, medical and clinical assessments occurred separately for males and females at their respective facilities.

For ethical reasons and to promote feasibility, recruitment and screening occurred as quickly as possible after admission, prior to the emergence of moderate-severity opioid withdrawal signs/symptoms. Screening included obtaining informed consent, HIPAA authorization, assessment of demographics, contact data, physical exam, pregnancy testing, contraception methods, medical and drug history, and urine drug testing.

During the recruitment and informed consent process, participants were repeatedly told they could receive FDA-approved MOUD at the treatment facility instead of participating in the study. They were also told that upon discontinuation of device stimulation, or if they dropped out of the study, they could receive MOUD at any time. Participants were told that device use (active or sham treatment) would be self-administered, that they could discontinue device use at any time and for any reason, and they could receive MOUD and comfort medications.

Patients admitted for OUD, between the ages 18–65 years old, in good general health, seeking to discontinue their illicit opioid use at the treatment facility, self-reporting that they wished to become abstinent without using MOUD, were eligible to participate. All participants provided informed consent and agreed to follow study procedures and use medically accepted highly effective contraception. Prior to receiving either active or sham treatment, all participants who enrolled had to exhibit at least moderate withdrawal severity, operationally defined as a total score of >13 on the Clinical Opiate Withdrawal Scale (COWS) (38). Exclusion criteria included pregnancy or lactation, serious current psychiatric disorder (schizophrenia, bipolar) or use of neuropsychiatric medications that may overlap with NET’s proposed mechanisms of action (e.g. anxiolytics, antidepressants, anticonvulsants, sedating histamine-1-receptor antihistamines, prescription or over-the-counter stimulants), need for detoxification from alcohol or benzodiazepines, past 300-day exposure to extended-release buprenorphine, certain chronic illnesses (especially seizures), unstable medical conditions, or presence of cardiac pacemaker.

All participants received the same clinical care as all patients in the facilities, which did not include planned psychosocial interventions following discharge. At discharge, participants were advised about their increased risk for opioid overdose (due to loss of opioid tolerance during the residential stay) and were provided with naloxone and trained in how to use it. The participants were not told whether they received active or sham intervention.

The NET Device delivers alternating current via surface electrodes placed transcranially (bilaterally) on the mastoid processes. The device delivers multiple low-amperage waveforms at controlled frequencies that vary throughout each treatment day, with no net direct current component.

In prior clinical studies, the output waveform was optimized for dynamic variations in skin impedance, electrode conductance, frequency and pulse-width related sensation, and orthogonal electrode pressure (e.g. from head pressure when sleeping), leading to improved rates of patient tolerability. Furthermore, human and animal studies, and clinical observation, have identified different electronic waveforms corresponding to subtypes of polysubstance use. These waveforms were programmed into the device at treatment entry based on each participant’s drug screen results at the time of treatment initiation. Stimulation was continuously available (except when bathing) for up to 7 days via transcutaneous electrodes of size approximately 1cm x 2cm. Stimulation output frequency varied from 4 to 3000 Hz and pulse width from 7 to 750 microseconds. Stimulation output current varied from 0 to 3.2 mA (peak) into a 15 kOhm load, and output voltage varied from 0 to 44 volts (peak to peak). Treatment was self-administered, and participants were instructed that they could control the device output intensity and duration according to perceived benefit. In prior studies, device output intensity has not been found to exhibit a significant relationship with withdrawal severity (NET Recovery Corp., unpublished data).

Sham treatment, to control for placebo effects, was designed to minimize treatment assignment recognition by the sponsor, principal investigator, independent study monitor, participants, research assistants, and treatment staff. The active and sham interventions both used the NET Device, but the sham intervention used lead wires that (although visually the same as active wires) were rendered non-conductive beforehand, preventing any electrical stimulation from being delivered to the participant.

Participants in the active and sham arms received identical instructions, equipment, electrode attachment methods and locations, and daily reviews of device operation. The apparatus presented both active- and sham-assigned participants with visual cues from the device’s “heartbeat” indicator (a blinking green light-emitting diode which indicated the device was active) during treatment. Research staff instructed each participant that the equipment was designed to be self-administered, that s/he could set the level of stimulation wherever it is comfortable, that stimulation is not always (and does not need to be) sensate, and that device use could be discontinued when the participant felt it was not providing additional benefit.

A research assistant used a 7-day timeline follow-back (TLFB) during a weekly video call to probe for substance use. TLFB reports were specific for MOUD (methadone, buprenorphine or naltrexone, which would be expected to lower opioid use), opioids without a prescription, psychostimulants (methamphetamine without a prescription, cocaine), and sedatives. These self-reported days of use were examined as percentage of total follow-up time (84 days).

Additional measures included demographic (sex, age, race) and clinical variables (length of residential stay, primary substance use disorder), randomization group, and duration of device use (obtained from internal device statistic in 5-min intervals). Based on the observed non-normal distribution of duration of device use, this variable was dichotomized as >24 hr and <24 hr of device use leading to four groups to maximize numbers in each subgroup: Active >24 hr, Active <24 hr, Sham >24 hr, and Sham <24 hr.

All participants who completed the trial (N = 108) were monitored during the follow-up period. Five participants who failed to complete a follow-up interview were not included in the analysis.

The primary analysis contrasted the groups randomized to active versus sham stimulation during the residential period for participation in post-discharge follow-up, completeness of data, and for outcomes using t-tests for continuous outcomes and chi-square for categorical outcomes. Subgroup analysis then dichotomized each randomization group by duration of device use (Active >24 hr, Active <24 hr, Sham >24 hr, and Sham <24 hr). When assumptions of normality were not met, nonparametric tests were used. Covariates included randomization group, duration of device use, and the demographic and clinical variables noted above.

Missing data for participants with any follow-up data were imputed as positive for substance use. In sensitivity analysis, Monte Carlo simulation was used to impute missing data based on observed means (39). Five passes were conducted with the outcomes averaged.

A total of 108 participants (53 active NET and 55 sham; 59.3% male, 89.8% white; 71.3% fentanyl-positive, and 50.0% psychostimulant [mostly methamphetamine] UDS at screening) completed the clinical trial and were eligible for the follow-up. Five participants randomized to the active device arm were lost to follow-up (3 incarcerated, 1 lost to follow-up, and 1 protocol violation treated as unblinded compassionate care), but no one in the sham device arm. They did not differ demographically or clinically from those who completed at least one follow-up interview. Table 1 presents their demographic and clinical characteristics at admission. Table 2 presents demographic and clinical characteristics of the four subgroups, who completed at least one post-discharge interview. The demographics of the four subgroups did not significantly differ, as we previously found for the two randomized groups (31).

The mean ± 1 standard deviation (SD) percentage of days using opioids or psychostimulants (with missing values for patient reports imputed as “positive for use”) for Active (36.3 ± 34.7%) versus Sham (39.1 ± 32.6%) over the total 84-day post-discharge period did not significantly differ (t = .425, p>.10). When comparing percentage use-days for the four subgroups over the entire post-discharge period, the Active >24 hr group reported numerically but non-significantly fewer mean use-days (1.3 ± 3.4%) than Active <24 hr (4.8 ± 10.8%), Sham >24 hr (4.7 ± 15.0%), and Sham <24 hr (4.8 ± 11.2%) groups, Kruskal-Wallis H(3)=0.87, p>.05.

Rates of concurrent weekly use of opioids and psychostimulants (i.e. using both drugs at least once during the same post-discharge week) were numerically but not significantly lower in the Active group (2.4% concurrent use-weeks) than Sham (4.2%, Fisher’s exact test p = .192, two-tailed). Rates of concurrent weekly use were numerically but not significantly lower in the Active >24 hr group (1.0%) than Active <24 hr (3.8%, Fisher’s exact test p = .055, two-tailed) and significantly lower than Sham >24 hr (6.8%, Fisher’s exact test p = .002, two-tailed), but not significantly different than Sham <24 hr (2.6%, Fisher’s exact test p = .314, two-tailed).

Given that post-discharge MOUD use could lead to decreased illicit opioid use, we first examined the percentage of MOUD use for the two randomized groups and four subgroups during the post-discharge period. The average percentage of days using MOUD (with missing values for patient reports imputed as “positive for use”) for Active (40.8 ± 34.2%) versus Sham (43.3 ± 33.9%) over the total 84-day post-discharge period did not significantly differ (t = -.367, p = .714). When comparing percentage use-days for the four subgroups, Active >24 hr reported numerically but non-significantly fewer mean use-days (0.7 ± 3.3%) than Active <24 hr (12.7 ± 23.6%), Sham >24 (6.0 ± 18.8%), and Sham <24 hr (9.6 ± 21.7%), Kruskal-Wallis H(3)=6.46, p = .087.

The average percentage of days using opioids or psychostimulants (with missing values for patient reports imputed as “positive for use”) for Active (44.5 ± 35.3%) versus Sham (48.9 ± 34.9%) over the total 84-day post-discharge period did not significantly differ (t = -.635, p = .526). However, the percentage of use-days significantly differed by subgroup, Kruskal-Wallis H(3)=10.05, p = .012. Pairwise comparisons using Mann-Whitney U tests indicated that the Active >24 hr group reported significantly less-frequent use (mean=2.0 days, median=0 days) than both Active <24 hr (mean=17.5 days, median=6 days, U = 138.5, z = -3.08, p <.05) and Sham <24 hr (mean=14.3 days, median=1 day, U = 259, z=-2.15, p <.05). The numerical difference relative to Sham >24 hr (mean=10.8 days, median=1 day, U = 160, z = -1.92, p = .055) was not significant. No significant differences were found between Active <24 hr, Sham >24 hr, and Sham <24 hr groups.

With missing data imputed as use, the four subgroups did not significantly differ overall, Kruskal-Wallis test H(3)=5.90, p = .12, although the Active >24 hr group reported numerically less-frequent imputed use (mean=27.8 days, median=11 days) than Active <24 hr (mean=46.2 days, median=50 days) and Sham <24 hr (mean=41.1 days, median=44 days).

Table 3 lists percentages of patients with zero days of opioid or psychostimulant use (complete abstinence) during the 84-day post-discharge period for the four subgroups. Table 4 lists summary statistics for these groups. Data are presented with reported use-days and with missing values imputed as use. The percentage of participants with zero days of use was significantly higher in the Active >24 hr group (73.9%) compared to the other groups (Active <24 hr: 32.0%, Sham >24 hr: 42.9%, Sham <24 hr: 47.1%), χ² = 8.97, p = .003.

The post-discharge group percentage of person-days with illicit opioid, psychostimulants, or MOUD use ( Table 5 ) was significantly lower in the Active >24 hr group (3.4% use-days) compared to Active <24 hr (31.6% use-days), Sham >24 hr (20.1% use-days), or Sham <24 hr (25.1% use-days), χ² = 370.10, 172.48 and 272.25, respectively, p <.001. Figure 1 presents the percentages of participants with any use of opioids, psychostimulants, or MOUD during each of the 3 months across the post-discharge period, to assess whether rates of use varied over time. Over the entire 84-day period, the proportion of participants using any illicit polysubstance or MOUD was significantly lower in the Active >24 hr group (26.1%) compared to Active <24 hr (68.0%, p = .004), Sham >24 hr (57.1%, p = .029), and Sham <24 hr (52.9%, p = .030, Fisher’s exact test).

Figure 2 illustrates, and Table 5 lists, percentages of post-discharge reported person-days separately for illicit opioid, illicit psychostimulant, or MOUD use in the four groups. Rates of opioid use were significantly lower in the Active >24 hr group (1.4% use-days) compared to Active <24 hr (7.0% use-days), Sham >24 hr (6.8% use-days), or Sham <24 hr (4.4% use-days), χ² = 51.33, 46.92 and 23.00, respectively, ps<.001, as were rates of psychostimulant use (Active >24 hr (1.3% use-days) compared to Active <24 hr (3.8% use-days), Sham >24 hr (6.7% use-days), or Sham <24 hr (4.3% use-days), χ² = 17.86, 49.49 and 24.63, respectively, ps<.001. Rates of MOUD use were also significantly lower in the Active >24 hr group (1.2% use-days) compared to Active <24 hr (23.0% use-days), Sham >24 hr (11.3% use-days), or Sham <24 hr (16.7% use-days), χ² = 299.37, 113.84 and 205.41 respectively, ps<.001, risk ratio from .106 to .052.

Figure 3 presents heat maps that depict individual-subject temporal patterns of post-discharge daily use of opioids, psychostimulants, and MOUD, and missing data, in each of the 4 residential treatment subgroups: (A) Active >24 hr, (B) Active <24 hr, (C) Sham >24 hr, and (D) Sham <24 hr. Individuals with higher quantities of non-use reports are clustered at the top and those with higher quantities of use reports are clustered at the bottom, facilitating visual comparison between the subgroups.

The Monte Carlo simulation results are consistent with analyses of reported use-days and missing data imputed as use. The Active >24 hr group had a significantly (χ² = 44.84, p <.001) higher percentage of participants with zero days of drug use (73.9%) compared to other groups (32.0–47.1%), and no participants with 84 days of drug use.

Figure 4 illustrates group-level daily percentage rates of TLFB reporting specific for all participants (4A), Active >24 hr (4B), Active <24 hr (4C), Sham >24 hr (4D) and Sham <24 hr (4E). Group-level retention did not follow a monotonically decreasing or stepwise survival function pattern, as many participants re-engaged even after periods of missing contact (56% to 74% on a given day). There were no significant differences in the number of days with no reporting between the Active and Sham groups as a whole (Mann-Whitney U = 1237.0, p = .59), or across the four groups (Kruskal-Walls H = .76, p = .69), Table 5 .

The randomized groups and four subgroups did not significantly differ by sex, age, or race. Duration of stay at the residential treatment facility was analyzed using non-parametric tests due to non-normal distribution of the data. The Active group duration of stay (mean=18.5 days, median=25.5 days, IQR = 7-27) was not significantly different from Sham (mean=16.6 days, median=17 days, IQR = 6-27, U = 1233.5, z = -0.57, p = .57). However, the distributions significantly differed by subgroup (Kruskal-Wallis H(3)=49.96, p <.001). Pairwise comparisons using Mann-Whitney U tests indicated that the Active >24 hr group had significantly longer residential stays compared to Active <24 hr (U = 142, p = .003), Sham >24 hr (U = 114, p = .003), and Sham <24 hr (U = 267, p = .044) ( Figure 5 ).

The correlation between residential stay duration (in 1-week bins) and whether participants used opioids, MOUD, or any of opioids, stimulants, or MOUD (binary: used vs. not used) during the post-discharge period (Spearman’s rho = .220, .246, and .064 respectively, ps>.05) was not significant.