This single-site, patient-assessor-blind, randomized, sham-controlled clinical trial was conducted in line with the Declaration of Helsinki. We followed the Consolidated Standards of Reporting Trials (CONSORT) (26) and Standards for Reporting Interventions in Clinical Trials of Acupuncture (STRICTA) (27) recommendations for designing and reporting of this trial. The protocol was reviewed and approved by the Human Research Ethics Committee (HREC) of the Shanghai Municipal Hospital of Traditional Chinese Medicine, Shanghai University of Traditional Chinese Medicine, Shanghai, China (No. 2020SHL-KY-42-02); and endorsed by the HREC of RMIT University, Melbourne, Australia (No. 24186). All participants were informed of the aims of this trial and randomization procedure giving them a 50% opportunity of receiving real-acupuncture (RA) or sham-acupuncture (SA), and all gave written informed consent at the time of enrollment in the trial. To increase potential participants' motivation to take part in the trial, it was explained that sham- and placebo- acupuncture has been reported to be associated with some positive outcomes [improvements in depression (28), and/or in insomnia (29)] in previous studies. All procedures including acupuncture treatment and outcome assessment were provided free of charge. This trial was registered at Chinese Clinical Trial Registry (ChiCTR) (Identifier: ChiCTR2100043054).

From February 2021 to December 2021, 70 participants were recruited at the Shanghai Municipal Hospital of Traditional Chinese Medicine via outpatient services, health promotion/education in the community, and advertisements on hospital-based posters and social media (WeChat, Tencent Holdings Limited).

Women were enrolled if they fulfilled the following criteria: (1) aged 45–55 years; (2) experiencing “perimenopause” defined according to the Stages of Reproductive Aging Workshop (STRAW) criteria (30); (3) met the International Classification of Diseases-Ten Edition (ICD-10) diagnostic criteria for mild or moderate depressive episode (31) and International Classification of Sleep Disorders-Third Edition (ICSD-3) diagnostic criteria for insomnia (32), and maintained comorbid depression and insomnia symptoms for at least 4 weeks; (4) 8 points ≤ 17 items-Hamilton Depression Rating Scale (HAM-D₁₇) ≤ 23 points; (5) Pittsburgh Sleep Quality Index (PSQI) >5 points; and (6) met the diagnostic criteria for “Depression of Liver and Deficiency of Kidney” pattern in Traditional Chinese Medicine (TCM) theory (see Appendix 1).

Participants were excluded if they: (1) showed non-natural menopause transition due to premature ovarian failure, medication intake, or surgery (e.g., oophorectomy); (2) were pregnant or currently lactating; (3) depression and/or insomnia were caused by systemic diseases (e.g., stroke, Parkinson's disease, surgery, etc.); (4) were diagnosed with a severe depressive episode according to ICD-10 criteria (31) and showed suicidal ideation/attempts; (5) had a depressive disorder diagnosis prior to perimenopause; (6) comorbid with other psychiatric and mental disorders (e.g., general anxiety disorder, bipolar disorder, obsession, etc.), other depression-related disorders (e.g., recurrent depressive disorder, dysthymia, etc.), or sleep-related disorders (e.g., obstructive sleep apnea syndrome, restless legs syndrome, etc.) diagnosed according to Diagnostic and Statistical Manual of Mental Disorders-Fifth Edition (DSM-V) (33) and ICD-10 (31) in addition to depressive episode and insomnia; (7) had any serious physical illness, such as serious diseases of cardiovascular or hematopoietic systems, or poor liver or kidney function; (8) took medications (e.g., antidepressant, hypnotic, other psychotropic substances, Chinese herbal medicine, etc.) and/or healthcare products (e.g., melatonin, black cohosh, etc.), or received other therapies (e.g., CBT, Mindfulness-Based Exercise, etc.) that are intended to treat depression/insomnia within the last 1 month, or received HRT within the last 2 months prior to the baseline; (9) had acupuncture treatment experience in the past 6 months; (10) had an infection close to the site of the selected acupoints; (11) alcohol or drug abuse/addiction; and (12) participated in any other clinical trial within last 3 months.

Participants in both groups received 17 sessions of acupuncture treatments for eight consecutive weeks (three sessions per week for the first 3 weeks, two sessions per week for the next 3 weeks, and one session per week for the final 2 weeks). All participants were provided with basic mental health and sleep hygiene education (see Appendix 2) before their first treatment session, which is the standard intervention given to all patients with climacteric-related mental illness at the trial site hospital.

Outcomes were appraised across four time-points, namely, baseline, end of the last treatment session (i.e., “post-treatment”), 8-week follow-up, and 16-week follow-up. The latter three time points were used to determine the short-, medium-, and long- term effects of acupuncture.

The projected sample size required was calculated on the basis of change of HAM-D₁₇ and PSQI global scores. In light of the previous RCTs (55, 56) with similar designs, the difference in HAM-D₁₇ scores and PSQI scores between women with PMD or PMI who received RA and those who received SA was 2.08 points and 3.73, respectively. The HAM-D₁₇-based between-group difference was smaller and was therefore selected for estimation of sample size to yield a more conservative result. PASS software (Version 15.0.5, NCSS, LLC) determined no < 31 patients needed to be recruited in each group to generate statistically significant results (α = 0.05, β = 0.1; the number of patients in the two groups are equal; Two-tailed test). Hence, a total of 70 participants were required for this trial with 35 in each group, allowing for a 10% attrition rate and other risks such as non-compliance, etc.

The random allocation list was generated by an independent research assistant (T-J) with no other role in this trial using SPSS software (IBM Corp., Armonk, NY, USA), and was prepared and concealed in 70 opaque, sealed, and sequentially numbered envelopes. The acupuncturist (H-X) opened the sealed envelopes once participants were at the first treatment session. If two or more participants received interventions on the same day, they were scheduled to come to the hospital at different times to avoid communication and interaction. Group allocation was kept confidential from participants, outcome assessors (FY-Z and YL-H), and the statistician (QQ-F). Only the independent research assistant (T-J) and the acupuncturist (H-X) knew the treatment allocation. The unblinding process was only executed after all the statistical analyses were completed.

Acupuncture in either RA or SA group was delivered by the same acupuncturist (H-X) registered in the Department of Psychiatry with over 30-years acupuncture clinical experience in treating psychosomatic diseases. Furthermore, the acupuncturist was asked to minimize interactions with participants to only brief social conversation and asking about AEs to avoid non-specific treatment effects. In addition, all relevant researchers received and passed the uniform, on-site, “Clinical Research Specification Training Program” prior to the commencement of this trial. Regular onsite monitoring was carried out by trial monitoring committee to ensure the interventions were delivered as planned and data collection was carried out and documented appropriately.

The continuous variables were described as mean ± standard deviation (SD)/mean ± standard error (SE) or median (inter quartile range), and the categorical variables were expressed by ratios or constituent ratios. For baseline characteristics, continuous variables were examined by independent sample t-tests or Mann–Whitney U tests, and categorical variables were assessed using Chi-square tests, or Fisher's exact tests as appropriate. The changes of efficacy-related outcome measures (HAM-D₁₇, PSQI, Meno-D, ISI, KI, MenQoL, HAM-A, and SCSQ) cross trial time-points were repeated measurement data. Paired-sample t-tests were adopted to address the between-group differences from baseline to each time-point; and linear mixed-effect models were adopted to detect main effect, time effect, and interaction (time^*group) effect. We utilized paired sample t-tests to compare the differences of changes in serum hormone levels, SRRS, AES, and MS-TSQ between two groups. Correlational analyses were adopted between HAM-D₁₇/PSQI and SRRS/SCSQ. All statistical analyses were carried out using SPSS software (IBM Corp., Armonk, NY, USA).

The James' (50) and the Bang's blinding indices (51) were analyzed using the Stata software (Version 17.0, StataCorp. LLC, College Station, TX, USA) with the blinding module.

All enrolled cases were analyzed by Intention-To-Treat (ITT) analysis, and the missing values were imputed using multiple imputation approach. We first used SPSS software to replace missing data with multiple (n = 5) imputations [predictive mean matching (PMM) model]; and then, the resulting F statistics were pooled using “miceadds:: micombine. F” in R software (Version 4.2.0, R Core Team), referring to the methodology of a previous RCT (57). However, as suggested by a previous paper regarding statistical methodology, when the proportion of missing values was more than 40%, Per-Protocol (PP) analysis should be used (58). Therefore, for those outcome measures which were assessed ≥3 time-points but the missing values were more than 40% in any time-point after baseline, we adopted both ITT analysis (using dataset obtained by multiple imputation) and PP analysis (using dataset including only remaining data after removal of missing values) to ensure the rigor of the results.

All tests in this study were two-tailed and the significance level was set to p < 0.05. All confidence intervals (Cl) were two-sided 95% intervals between the two groups.

A total of 117 participants were screened for eligibility and 47 were excluded. The remaining 70 participants were randomly assigned to receive RA (n = 35) or SA (n = 35). Of these, 65 (92.9%) completed all 17-week treatment sessions, 54 (77.1%) completed the 8-week follow-up visit, and 41 (58.6%) completed the 16-week follow-up. Reasons for withdrawal are displayed in the flow diagram (Figure 1), and the numbers and reasons are similar between the two groups.

Tables 1, 2 demonstrate the socio-demographic and clinical characteristics of the study participants. Sixty-six (94.3%) participants were married or cohabiting, and 64 (91.4%) had a secondary school diploma/degree or higher. There were no alcohol users among the participants, whereas some had a daily coffee/tea drinking habit. The mean age of participants was 48.9 years old. The demographic factors and baseline outcome assessments were comparable between the RA and SA groups (all p > 0.05).

Twenty-nine participants (41.4%) withdrew from this trial at the 16-week follow-up. We imputed missing data per description in the methods. For those outcomes involving assessments at follow-ups, we performed both t-tests and linear mixed-effect models analysis using the ITT dataset (data of all 70 participants) and PP dataset (data of 41 participants who complete all assessments), respectively. The results derived from both datasets are displayed, compared and analyzed as follows. For the outcomes only assessed at baseline and post-treatment (withdrawal participants = 5, 7.1%), t-tests were used.

Each participant was asked to rate their experience of needling sensations using the MASS scale. Sixty-six participants (34 in RA group; 32 in SA group) who completed at least nine treatment sessions completed this appraisal (Table 9). Thirty-three (97%) participants in the RA group experienced mild to moderate soreness and 14 (41%) experienced fullness/distension. Whereas, in the SA group, sharp pain was the main sensation experienced by 22 (69%) participants; and 10 participants (31%) also experienced aching. We further analyzed the intensity of each sensation using the Mann–Whitney U test (Table 10). There were significant differences in soreness, fullness/distension, numbness, and sharp pain (all p < 0.01), with the RA group more likely reporting a higher degree of soreness, fullness/distension, and numbness whereas the SA group reporting a higher degree of sharp pain.

Change in participants' expectation of acupuncture from baseline to the end of end of ninth treatment session was recorded by AES scale. Both groups showed a significant downward trend in AES scores (both p < 0.05), although no significant between-group difference was detected [0.07 (95% CI, −0.61 to 0.76), p = 0.83] (Table 11).

Analysis of PP dataset was unnecessary for MS-TSQ given only four and only five participants did not provide data at end of ninth treatment session and post-treatment assessments, respectively. As shown in Table 12, at post-treatment, the satisfaction scores of RA group were significantly higher than that of the SA group in three dimensions only (sleep quality, concentration, and overall) (all p < 0.05). The mean value of overall satisfaction scores in RA group was 3.29, between “Satisfied” and “Extremely Satisfied,” while the mean value of overall satisfaction in SA group was 2.28, between “Neutral” and “Satisfied,” according to scoring principles of MS-TSQ scale (54). As treatment progressed, the satisfaction scores of RA group increased significantly in sleep quality and overall dimensions (both p < 0.05), and such changes were superior to the changes in SA group [1.02 (95% CI, 0.61–1.43), p < 0.01 in sleep quality dimension; 0.86 (95% CI, 0.49–1.24), p < 0.01 in overall quality dimension].

In the RA group, four (11.43%) participants reported pain at needling sites, and two of them also reported subcutaneous hematoma around needling sites. In SA group, two (5.71%) participants reported dizziness, and one of them also reported pain at needling sites; another (2.86%) reported numbness in the arms. All AEs were considered mild by participants. Moreover, according to participants' self-reports, these AEs usually appeared after the needles were removed and gradually disappeared over the following 24–72 h. None of the participants withdrew due to AEs or required medical attention.

Sixty-five participants who received all treatment sessions completed the APS appraisal, in which participants were asked to make a guess of their group allocations as well as the basis for their guess. As exhibited in Table 13, a majority of participants in either group judged the type of acupuncture treatment they received based on its clinical effectiveness.

The James' blinding index was 0.67, and the upper bound of the 95%CI was 0.75 (>0.5), suggesting there was insufficient evidence to show unblinding between the two groups. Furthermore, although two thirds of the participants (67%) in the RA group correctly guessed their group allocation, comparing to less than half (44%) in the SA group, the Bang's blinding index of the SA group was-−0.14 (−0.20 < Bang's blinding index cut-offs < 0.20), suggesting that participants in SA were also successfully blinded (Table 13). Taken together, we can assume that blinding was successful in our study and the efficacy-related outcomes thereby were credible.

To the best of our current knowledge, this is the first sham-controlled RCT to investigate acupuncture for comorbid depression and insomnia during perimenopause, and to assess the effect in the short-, medium-, and long- term. Blinding was successful and only minor AEs were reported. The RA group reported significantly reduced PSQI scores at post-treatment and 8-week follow-up, to a greater extent than the SA group, but did not differ from the SA group on either the reduction of HAM-D₁₇ scores or changes in serum reproductive hormones. Overall, acupuncture is safe and contributes to clinically relevant improvements in insomnia among women with comorbid PMD and PMI, with satisfactory short- and medium- term effects.

There are a few strengths of the current trial. First, to ensure we recruited women with similar presentation, we not only selected perimenopausal women with comorbid PMD and PMI, but also limited recruitment to one specific type of TCM syndrome diagnosis “Depression of Liver and Deficiency of Kidney,” which is the most common pattern of presentation confirmed by bibliometric studies (23). This design mirrors the clinical scenario, in which treatment regimen should be tailored to individual syndrome pattern or syndrome cluster (70). By limiting to people with one specific syndrome pattern we were able to semi-standardize our trial treatment and to enhance internal validity. Second, we used a SA procedure that is invasive with minimal needle manipulation to acupoints that have no specific effect on either insomnia or depression (35) to enhance the blinding. This SA design has been used in an acupuncture trial for insomnia with successful blinding of participants (85), and it was verified with both James' and Bang's blinding indices in our trial. This choice of SA is due to many Chinese women would have some knowledge about acupuncture. Non-invasive needling to non-acupoints would compromise the blinding. Third, it was thought that female estrogen might explain the acupuncture effect on PMI (6, 36). Our study was the first to examine this hypothesis and found reproductive hormone changes do not explain the improvements. Finally, we applied ITT and PP analyses and found the results were consistent, suggesting the robustness of the findings.

Due to its preliminary nature, some limitations of this trial should be acknowledged. First, larger than anticipated cases lost to follow-up forced us to use multiple imputation approach to replace the missing data. Started in March 2022 and lasted for more than 3 months, Shanghai witnessed the China's worst outbreak of SARS-CoV-2 virus since early 2020 (86). Seven participants declined further follow-ups due to this sudden outbreak, and reported that the pandemic exacerbated their mood and sleep disturbances. It is undeniable that there might be subtle differences between these software-generated data and the real experience of the participants. However, such difference does not appear to significantly impact on our findings, given the results of ITT and PP were consistent in all the key outcome measures. Second, the long-term efficacy of acupuncture might have been somewhat underestimated because five participants took part in the 16-week follow-ups during this strong wave of pandemic. It is likely they too experienced worsened sleep or mood during this time as those who dropped out during this period. Third, whilst self-reported sleep has improved following RA intervention, if and how acupuncture might change sleep process and architecture is unknown as we did not include PSG in the trial. We made this decision due to previous experience at the study center. After the initial Wuhan outbreak of COVID-19, almost all participants who had been interested in sleep clinical trials conducted at our center declined in-hospital PSG sleep monitoring. Considering the participants' willingness and the trial's feasibility, we did not include PSG in the current RCT. Fourth, without PSG, the screening of other sleep disorder, such as sleep apnea or restless leg syndrome, was based on trial doctor's thorough inquiry, diagnostic criteria for sleep disorder (i.e., DSM-V and ICSD-3), and participants' medical history; it is therefore possible some participants had more than PMI. This is, however, not likely to affect our current conclusion regarding the application of acupuncture to women with simple PMI. Comorbid insomnia and other sleep disorders can only be more complex and difficult to treat than simple insomnia. Even if the included participants incorrectly included those with comorbid PMI and other sleep disorders, we have affirmed the sleep-promoting effect of acupuncture in them. Fifth, participants with severe depressive episode/insomnia or other serious physical/mental illness were excluded at screening stage. Our findings thereby cannot be generalized to sicker patients, or perimenopausal women accompanied with other complex/severe comorbidities.

The current trial has provided essential data on safety, acceptance and effectiveness of acupuncture for comorbid PMI and PMD. Future trials may consider a wait-list control, expanding the sample sizes, and extending the follow-up period, to confirm the current findings and clarify whether the remission in depression is a specific effect of acupuncture or only a mega-placebo or natural regression. It is also necessary to examine when acupuncture is combined with sedating-antidepressant agent if it can further augment anti-depressive/sleep-promoting effects or further reduce medication-induced AEs. Collecting the information regarding participants' previous acupuncture experience is also encouraged in future study since this experience may affect their acceptability and compliance.

Clinically, comorbid PMI and PMD may require multidisciplinary approaches (8, 9, 23), and acupuncture could be an ideal adjunct therapy for insomnia and depression. A sufficient number of acupuncture treatment sessions is required for the optimal effect. Given we were interested in assessing the short-term, medium term and long-term effects of acupuncture and did not want to burden participants with data collection at multiple points, we did not track participants' weekly or monthly changes in PMI and PMD during the 8-week treatment. We therefore cannot comment on the progression of participants or the accumulative effect of acupuncture. Future studies may look into the weekly changes immediately after acupuncture treatment to gain a better understanding on this aspect.