Literature search was conducted in eight databases, namely, PubMed, Embase, Cochrane Library, Web of Science (WoS), China National Knowledge Internet (CNKI), Wanfang Data, VIP Database for Chinese Technical Periodicals, and China Biology Medicine disc (CBM). The search period was from their inception to 15 March 2024. The main search terms include somatic symptom disorder, acupuncture, and randomized controlled trial. The search strategy was compiled by combining free words and subject headings and tailored according to the characteristics of each database. The references of previous reviews were also searched for relevant literature. The detailed search strategies are described in Supplementary material S1.

Studies that met all of the following criteria were included in this review: (a) patients were diagnosed as “somatic symptom disorder” or “bodily distress disorder (BDD)” according to the Diagnostic and Statistical Manual of Mental Disorders, 5th Edition (DSM-5 TM) or International Classification of Diseases 11th Revision (ICD-11), regardless of age, sex, and severity of the disease; (b) at least one group of interventions in the study was acupuncture alone or acupuncture in combination with other conservative or pharmacological treatments; (c) the type of study design was randomized controlled trial; (d) outcomes included, but not limited to, Hamilton Depression Scale (HAMD), the Hamilton Anxiety Scale (HAMA), Symptom Checklist-90 (SCL-90), etc.; (e) the language is limited to Chinese or English.

Studies with one of the following characteristics will be excluded: (a) patients had other types of mental or psychiatric disorders, or were lactating and pregnant women; (b) herbal medicine or moxibustion was included in any intervention arms, or comparison of interventions was made between different types of acupuncture, or different frequencies or protocol of the same type of acupuncture; (c) systematic reviews, secondary analyses of RCTs, and conference proceedings; (d) no required outcome data available for data analysis.

After removal of duplications, titles and abstracts of the identified studies were independently reviewed by two reviewers (FZ and WW) against the inclusion and exclusion criteria of the review. Studies that were not relevant to the review were excluded. Full texts of the remaining studies were subsequently retrieved and reviewed by the two reviewers to identify eligible studies. Any disagreements were solved by senior reviewer (YL or HC).

The data were extracted from included studies according to a standardized form with the following information: author, year, country, diagnostic criteria, demographic characteristics of target population, intervention regimens of the experiment and control groups, time of follow-up, outcome measures, methods of statistics, and results. In addition, the number, type, and severity of adverse events were recorded. Two independent reviewers conducted the data extraction, and disagreements were completely discussed or solved by a senior reviewer.

The risk of bias assessment was conducted using Cochrane Risk-of-Bias (RoB) tool version 2. The risk of bias of included trials was assessed in each of the five domains, and an overall assessment of “high risk”, “some concerns”, or “low risk” was rated. The major domains of bias were as follows: randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result. Two reviewers independently assessed the risk of bias of included studies using RoB2. Any disagreement was resolved by discussion, with involvement of a third reviewer where necessary.

The Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach was conducted to evaluate the certainty of evidence.

Continuous variables were presented as mean difference (MD) with standard deviation (SD), and dichotomous variables by risk ratio (RR) with 95% CI, respectively. Meta-analysis was conducted to present a pooled estimate of the treatment effect where similar study designs, intervention, and outcome measures were identified among studies. A fixed effect model was adopted for studies when heterogeneity was acceptable (I² ≤ 50%), and a random effect model adopted when substantial heterogeneity was detected (I²>50%) in meta-analysis. Narrative analysis was performed for the data that could not be synthesized statistically. The RevMan V.5.4 was used for data analysis and synthesis.

The eight electronic databases were searched, a total of 9,526 articles were retrieved, and 1,383 duplications were removed. A total of 8,092 articles were excluded after screening by title and abstract according to eligibility criteria. Finally, 51 articles were screened by full-text reading, of which 46 studies were excluded, and 5 studies were eventually included. Figure 1 illustrates the study selection process.

The five studies included were all from China with a total of 376 patient samples (22–26). The studies by Chen (22), Sun et al. (23), Ma et al. (25), and Ren et al. (26) focused on patients with persistent somatoform pain disorder (PSPD, a subtype of SSD), while the other study (24) targeted general SSD patients. All studies reported that the baseline data were comparable. Four of the studies (22–25) used different types of acupuncture combined with antidepressants as the experimental group and antidepressants alone as the control group to evaluate the incremental effect of acupuncture for SSD patients. One study (26) used floating acupuncture combined with a placebo as the experimental group and simulated floating acupuncture combined with antidepressants as the control group. The treatment duration ranged from 6 to 8 weeks, and the reported outcomes included the Hamilton Anxiety Scale (HAMA), Hamilton Depression Scale (HAMD), numeric rating scale (NRS), the Symptom Checklist-90 (SCL-90), Pittsburgh Sleep Quality Index (PSQI), World Health Organization Quality of Life (WHOQOL), and Short-Form McGill Pain Scale (SF-MPQ), including Pain Rating Index (PRI), Visual Analog Scale (VAS), and Present Pain Intensity (PPI) (shown in Table 1).

In terms of randomization process, Chen (22) and Sun et al. (23) mentioned the use of random number table method for randomization in the original article. The remaining three studies (24–26) only mentioned randomization without describing specific methods, so it was considered that these three studies may be randomized, but because there was no significant difference in baseline characteristics between the two groups of participants in the three studies, they were assessed as low risk.

Due to the nature of acupuncture, blinding is often difficult to implement in non-sham acupuncture-controlled trials. One study (26) used simulated floating acupuncture in the control group to blind the patients, but the acupuncturist could not be blinded, so the study was rated as some concern in terms of the deviation of intervention measures. The rest four studies (22–25) adopted medication only in the control group, and it was not possible to blind both patients and treatment providers; therefore, they were rated as high risk of bias under this domain.

For missing outcome data, no study reported missing data; therefore, low risk of bias was rated. For selection of the reported results, protocols of all five studies were not available publicly, so we were not sure whether there were unreported outcomes, and some concerns were rated.

For the measurement of the outcome, the assessment criteria of Liang and Liang (24) were not clear, and the outcome measures used in the study were not specific for SSD and were assessed as some concern. Therefore, the overall risk of bias for all studies was rated as some concerns (shown in Figures 2, 3).

Three outcome measures were included in the GRADE assessment, and every outcome was reported by two trials. For HAMA scores, the evidence was rated as low certainty at 2 and 4 weeks (downgraded due to serious risk of bias and imprecision) and very low certainty at 6/8 weeks (further downgraded for inconsistency and indirectness). For NRS scores, evidence across all timepoints (2, 4, and 6/8 weeks) was assessed as very low certainty (downgraded for serious risk of bias, inconsistency, and imprecision). Adverse event data demonstrated moderate certainty (downgraded once for risk of bias). The full GRADE summary of findings is presented in Supplementary material S2.

Of the included studies, four compared the effect of acupuncture combined with antidepressant [selective serotonin reuptake inhibitor (SSRI) and serotonin and norepinephrine reuptake inhibitor (SNRI)] to antidepressant alone. Chen (22) (n = 82) compared acupuncture combined with paroxetine (SSRI) to paroxetine alone; Sun et al. (23) (n = 60) compared electroacupuncture combined with duloxetine (SNRI) to duloxetine alone; Liang and Liang (24) (n = 70) compared acupuncture combined with duloxetine (SNRI) to duloxetine alone; Ma et al. (25) (n = 94) compared floating acupuncture combined with duloxetine (SNRI) to duloxetine alone (shown in Table 1).

The study of Ren et al. (26) (n = 70) compared the effect of floating acupuncture combined with placebo to simulated floating acupuncture combined with duloxetine in patients with SSD for 6 weeks.

The results showed that the difference in change of HAMA from baseline between the experiment and control groups at week 1 (−2.30, 95%CI: −3.41 to −1.19; p = 0.0001) was statistically significant; however, the difference was not significant at weeks 2 (−0.02, 95%CI: −1.26 to 1.22; p = 0.97), 4 (−0.77, 95%CI: −2.12 to 0.58; p = 0.27), and 8 (−0.05, 95%CI: −1.74 to 1.64; p = 0.95), respectively. Similarly, the difference in change of HAMD from baseline was statistically significant between the experiment and control groups at week 1 (−2.39, 95%CI: −3.50 to −1.28; p = 0.0001) but not significant at weeks 2 (−0.14, 95%CI: −1.34 to 1.06; p = 0.82), 4 (−0.70, 95%CI: −2.03 to 0.63; p > 0.31), and 8 (−0.18, 95%CI: −1.07 to 1.34; p = 0.82) after treatment (shown in Table 2).

SF-MPQ (PRI, PPI, and VAS) was used to evaluate the effect of floating acupuncture for PSPD patients. The difference in change of PRI (−1.11, 95%CI: −2.20 to −0.02; p = 0.05) and PPI (−0.63, 95%CI: −0.86 to −0.40; p = 0.0000) between the two groups was statistically significant after treatment, indicating a better effect on pain relief in the experiment group. The difference in change of VAS score (−0.72, 95%CI: −1.19 to −0.25; p = 0.004) and PPI score (−0.39, 95%CI: −0.65 to −0.13; p = 0.004) was statistically significant between the two groups at week 1 after treatment but was not significant between the two groups at other time points (p > 0.05) (shown in Table 2).

Of the five studies included, only three studies reported adverse effects. Chen’s (22) and Ren et al.’s (26) studies reported adverse events in the form of Treatment Emergent Symptom Scale (TESS) score, while Ma et al. (25) reported adverse effect through self-report (shown in Table 2).

Chen (22) reported that, in the early stage of study, there were two cases of nausea, one case of drowsiness, and one case of dry mouth in the experiment group, while there were two cases of nausea, two cases of anorexia, one case of drowsiness, and one case of constipation in the control group. After 8 weeks of treatment, the above adverse events were alleviated or disappeared.

Ma et al. (25) reported nausea and vomiting, dry mouth, constipation, and decreased appetite. However, there was no significant difference in the incidence of adverse events between the two groups (Fisher’s exact test, p = 0.57).

Ren et al. (26) reported 3 cases (8.3%) of adverse events in the experiment group and 17 cases (50.0%) in the control group, and difference in the incidence of adverse events between the two groups was statistically significant (Fisher’s exact test, p = 0.0001). Similarly, most of the adverse events, which were mild or moderate, occurred in early stages of treatment and alleviated or disappeared later. No serious adverse events were observed.

There are some limitations of this study that could affect the interpretation of the result: (1) the number of studies included in this review was very small with small sample size, and the overall quality of the studies included in this review was moderate to low; this largely undermined the confidence in the result; (2) due to absence of SSD treatment guideline or consensus, and characteristic of acupuncture practice, there is large heterogeneity of outcome indicators, acupuncture type and treatment protocol, study design, etc., which limited the synthesis of outcomes data; (3) follow-up periods of all studies were short, and the long-term effects could not be observed; (4) all included studies were from China, which may limit the extrapolation of the results; (5) the limited number of studies and their intervention design precluded further analyses to distinguish the effect between acupuncture combined with SSRIs and SNRIs in this review.