This clinical trial was designed as a prospective, double-blind, randomized controlled study and will be conducted at the Affiliated Hospital of Yangzhou University. A total of 165 patients will be recruited based on stringent inclusion and exclusion criteria. All participants will provide written informed consent before enrollment (Supplementary Material 1). The study adheres to the Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) guidelines (20). Table 1 shows the SPIRIT schedule of enrolment, intervention, and assessments. The study flow diagram is shown in Figure 1.

Study participants will be randomly allocated to one of three groups (group A: active tDCS + ZOL, group B: active tDCS + placebo, group C: sham tDCS + ZOL) in a 1:1:1 ratio based on computer-generated random numbers made by an independent research staff who is neither involved in the enrolment nor in the assessment of the participants. Each participant will receive a unique, de-identified alphanumeric study code that records their enrollment order and allocation group, ensuring data traceability and confidentiality. The randomization assignments will be safeguarded in sequentially numbered, opaque, sealed envelopes by a designated research coordinator. Immediately before the patient enters the treatment room, a nurse will unseal the envelope to confirm the assigned group. After verifying the details of group allocation, the envelope will be resealed, and this step will be carried out without revealing the information to the patients. To maintain blinding, participants will be advised during the informed consent process that possible sensory perceptions (such as pricking sensations, needle-like stimulation, warmth, and vibration) may occur during treatment and that their intensity will be independent of therapeutic effectiveness. Additionally, all stimulation devices are identical in appearance, with differences restricted to electrical output parameters, and the ZOL and placebo pills have the same size, appearance, weight, and odor. This matching ensures the integrity of the blinding procedure. Blinding success will be evaluated once, at the end of the 4-week intervention period, using a standardized questionnaire for both participants and intervention providers. Participants and intervention providers will be asked to guess their own group assignment and to rate their confidence in that guess. Throughout the study, all outcome assessors, data analysts, intervention providers, and study participants will remain blinded to group allocation. Outcome assessors will additionally undergo rigorous blinding training prior to the study commencement to standardize assessment procedures and minimize subjective bias.

Should a serious adverse event (SAE) or other emergency occur during the trial, immediate unblinding will be performed to enable appropriate medical intervention. The decision will be a collaborative one between the principal investigator (PI) and the attending physician. Each unblinding event will be fully documented in the Case Report Form (CRF), including the reason for unblinding, date, time, personnel involved, and subsequent actions taken.

The active tDCS + ZOL (combined) group will be treated with tDCS using a tDCS instrument and take 10 mg zolpidem 10–30 minutes before bedtime nightly. The stimulator model is E-TDCS03 (Shenzhen Ailite Medical Technology Co., Ltd., Shenzhen, China), which provides 2 mA of stable direct current power through a battery. Two sponges (5 × 5 cm², soaked in 0.9% normal saline to ensure conductivity) will be placed on the rubber electrodes for stimulation. The left (F3) and right (F4) dorsolateral prefrontal cortex (DLPFC) regions, as defined by the International 10–20 EEG Electrode Distribution System (21), will be the sites for the anode and cathode electrode placement, respectively. A fade-in/fade-out design (30 s each) will be used to decrease potential skin sensations during the beginning and end of the stimulation (22). During tDCS intervention sessions, patients will be requested to sit in a comfortable reclining chair, switch off their smartphones, and are encouraged to hydrate, relax, or, and fall asleep. Communication with nursing staff will be minimized. The tDCS intervention will be administered according to a fixed protocol: 20-minute sessions, a fixed time during the daytime once daily from Monday through Friday, over four consecutive weeks. This regimen will be delivered by two trained nurses.

The active tDCS + placebo group will be treated with tDCS and take a placebo pill 10–30 minutes before bedtime nightly. The stimulation modality, stimulation electrode positions, stimulation region, and stimulation duration will match those adopted in the combined group.

The sham tDCS + ZOL group will undergo treatment with pseudo-stimulation using tDCS and take 10 mg zolpidem 10–30 minutes before bedtime nightly. The stimulation modality, stimulation electrode positions, stimulation region, and stimulation duration will be identical to those applied in the combined group. The pseudo-stimulation will only output a current of up to 2 mA for 10 seconds at the start and end of the session, allowing participants to perceive a sensation comparable to real stimulation. The device will cease delivering electrical stimulation throughout the middle of the 20-minute session.

During the entire intervention period, researchers with professional training will strictly abide by the protocol and closely track participants’ responses to prevent adverse reactions. Should participants report unbearable discomfort, the intervention will be terminated immediately.

The primary outcome measure is the response rate at 4-week follow-up (time point of finishing 4-weeks treatment period), defined as the percentage of those having at least a 50% reduction in insomnia symptoms from baseline as measured via the PSQI (9). The PSQI is a 19-item questionnaire comprising seven components: sleep quality, sleep latency, sleep duration, sleep efficiency, sleep disturbance, use of sleep medication, and daytime dysfunction. Each component is rated from 0 to 3, yielding a global score between 0 and 21, where a higher total score denotes worse sleep quality. This scale is widely employed to assess participants’ subjective sleep quality over the preceding month (5, 23).

Potential adverse events (AEs) related to electrical stimulation, including skin redness, pruritus, headache, epileptic seizure, hypomania, and common adverse events such as dizziness, or nausea, will be closely monitored. Any AE will trigger a prompt clinical evaluation and necessary medical care. Mild AEs (e.g., minor discomfort) will be managed with localized symptomatic treatment. In cases of moderate or severe AEs (e.g., syncope, severe pain), the intervention will be immediately halted or discontinued, and the participant will be continuously monitored by qualified medical personnel until complete resolution. Comprehensive emergency protocols are established to ensure unwavering patient safety throughout the trial. All AEs will be systematically recorded in the CRF with a comprehensive assessment of their causal relationships.

Sample size was calculated using PASS 15.0 software. Response rates at the 4-week follow-up were 53.3% (8/15) in the active tDCS + ZOL group, 20% (3/15) in the active tDCS + placebo group, and 13.3% (2/15) in the sham tDCS + ZOL group in our pilot study. We conservatively assumed a 50% response rate at the 4-week follow-up in the active tDCS + ZOL group, a 20% response rate in the active tDCS + placebo group, and a 15% response rate in the sham tDCS + ZOL group, thus a sample size of 44 per group is required to have a power of 80% and a two-tailed alpha level of 5%. Considering the 20% attrition rate and a block size of 3, we set our enrolment target at 165 subjects (55 per group).

Data analyses will be performed using SPSS 26.0 (IBM, Armonk, NY, USA) and GraphPad Prism 10.0 (GraphPad Software, San Diego, CA, USA). The normality of continuous data will be assessed using the Kolmogorov-Smirnov test, and the homogeneity of variances will be examined with Levene’s test. Continuous variables will be presented as mean ± standard deviation if normally distributed, and as median with interquartile range (IQR) otherwise. Between-group differences will be evaluated using one-way ANOVA for normally distributed continuous variables, the Kruskal-Wallis test for non-normally distributed continuous variables. Chi-squared test will be used to evaluate the between-group differences among the three groups in categorical variables. Time-to-event outcomes will be assessed using Kaplan-Meier curves and Log-rank test. Multivariate analyses will utilize multiple linear or logistic regression models based on dependent variable type. To analyze treatment effects, repeated measures data will be analyzed using generalized estimating equations (GEE). This analysis will explore the changes in clinical outcome data across different treatment groups and timepoints. The treatment-by-time interaction will be tested first. If this interaction is statistically significant, between-group differences will be assessed at each individual time point with a Bonferroni correction for multiple comparisons. Otherwise, the main effect of treatment will be tested without applying the Bonferroni correction. The primary analysis will follow the intention-to-treat (ITT) principle, with missing data handled by multiple imputation. A per-protocol analysis will be performed as a sensitivity analysis. A two-sided p-value < 0.05 will be considered statistically significant for all tests.

The rs-fMRI images will be acquired using a 3.0 T Ingenia MR scanner from Philips in Amsterdam, Netherlands, equipped with a 32-channel birdcage head coil. All participants will undergo fMRI scanning once before the start of treatment and again at the completion of the 4-week treatment period. During scanning, participants will be instructed to remain motionless and awake, with their eyes closed and minds clear of any thoughts. Functional images will be acquired with the following parameters: field of view (FOV) = 240 × 240 mm, matrix = 64 × 64, repetition time (TR) = 2000 ms, echo time (TE) = 30 ms, slice thickness = 3.5 mm, interslice gap = 0.25 mm, 38 axial slices in parallel, and 240 time points. High-resolution T1-weighted anatomical images will also be captured using the parameters: TR = 10 ms, TE = 4.7 ms, flip angle = 9°, FOV = 256 mm × 240 mm, matrix = 320 × 300, and slice thickness = 0.8 mm.

The fMRI data will be preprocessed using MATLAB-based CONN 18b toolbox within SPM 12. Preprocessing of the whole brain will include correction of slice timing, image realignment, structural segmentation, spatial and functional normalization to a standard MNI space, resampling into 2×2×2 mm voxels, and application of a 6 mm full-width half-maximum (FWHM) Gaussian kernel smoothing. Subjects will be excluded from the data analysis if their mean framewise displacement (FD) exceeds 0.2 mm, to reduce the impact of head motion. The ART feature from the CONN toolbox will be utilized (https://www.nitrc.org/projects/artifact_detect/) to identify outlier time points for movement parameters and overall signal intensity. Potential variables like white matter, cerebrospinal fluid signal, linear trend, subject motion (six rotational/translational motion parameters and six first-order time derivatives), and outliers will be included as regressors in the linear regression analysis. Next, the remaining blood oxygenation level dependent (BOLD) signal will be processed using a band-pass filter (0.008-0.09 Hz) to reduce the impact of low-frequency fluctuations and high-frequency interference.

Following preprocessing, fALFF and ReHo maps will be calculated for each participant. To standardize these metrics for group-level comparison, individual fALFF and ReHo maps will be converted to z-scores by subtracting the global mean and dividing by the global standard deviation. Seed-based FC analysis will be performed using CONN 18b. The left and right dorsolateral prefrontal cortex (DLPFC) will be defined as spherical regions of interest (ROIs) based on probabilistic maps from the SPM Anatomy toolbox (28). At the first-level (individual) analysis, the mean BOLD time series will be extracted from each of the bilateral DLPFC seeds for every participant. Subsequently, Pearson’s correlation coefficients will be computed between each seed’s time series and the time series of all other voxels in the brain, generating individual-level whole-brain correlation maps for each seed. These correlation maps will then be converted to Z-score maps using Fisher’s z-transformation for subsequent group-level analysis. At the second-level (group) analysis, the z-standardized fALFF maps, ReHo maps, and FC Z-score maps will be analyzed using GEE. Statistical significance will be assessed using a cluster-forming threshold of p < 0.001 at the voxel level, with cluster-level inferences corrected for family-wise error (FWE) at p < 0.05.