We performed the systematic review in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (18). We searched PubMed, Embase, Cochrane Library, Web of Science, China National Knowledge Infrastructure (CNKI), Wanfang Database, Chinese Scientific Journal Database (VIP), and Chinese Biomedical Literature Database (CBM) databases from inception to 23 June 2023 and identified randomized controlled trials (RCTs) examining the use of acupuncture in patients experiencing fatigue, depression, anxiety, cognitive dysfunction, headache, and insomnia, provided these symptoms were not caused by other chronic conditions. The selection of these symptoms was based on published studies that reported the common symptoms of long COVID (8, 9, 19, 20). We also carried out backward and forward searching to identify additional studies from citations. The completed search formula for each database is depicted in Supplementary Table 1.

Detailed inclusion and exclusion criteria are listed in Supplementary Table 2. RCTs reporting patients with at least one of the targeted neurological or neuropsychiatric symptoms (fatigue, depression, anxiety, cognitive abnormalities, headache, and insomnia) were included. Studies were excluded if the symptoms were caused by other chronic disorders.

Two pairs of reviewers (MXYL and YZ, DW and WCL) reviewed the title and abstract of each article and selected studies for a full-text review based on the inclusion and exclusion criteria. Conflicts were resolved through discussions and the involvement of a third researcher (LLZ).

Three reviewers (HL, WCL, and MXYL) extracted participant characteristic data from the selected studies using a standardized data extraction form. We extracted the following information from each included article: first author, year of publication, type and duration of the diseases, diagnostic criteria for the neurological and neuropsychiatric symptoms, number of participants, demographic data of participants, characteristics of the acupuncture and control groups, outcome data, duration of follow-ups, number of dropouts, and reports of adverse effects. Given that the number of participants and dropouts might vary at different stages in the same study, we considered the sample size included in the intention-to-treat analysis and the number of dropouts at the end of interventions.

All reviewers underwent comprehensive training for conducting assessments using the guidance document developed by the Cochrane Risk-of-Bias Tool for Randomized Trials (RoB 2) Development Group (21). The training encompassed a detailed understanding of the assessment process, including the nuances of each bias domain and the application of the tool's criteria. Two reviewers (DW and MXYL) assessed the potential risk of bias for all included studies in five different domains: (1) bias arising from the randomization process; (2) bias due to deviations from intended interventions; (3) bias due to missing outcome data; (4) bias in the measurement of the outcome; and (5) bias in the selection of the reported result. Each domain was assessed for risk of bias as “low risk,” “some concerns,” or “high risk,” and an overall consensus about the risk of bias was reached. In cases where a high risk of bias was identified, detailed documentation of the particular concerns within each domain was undertaken. A third review author (WCL) acted as an adjudicator in the event of disagreement.

We conducted a systematic review and meta-analysis for the following symptoms: fatigue, depression, anxiety, mild cognitive impairment, headache, and insomnia. For the meta-analysis, mean differences (MDs) with 95% confidence intervals (CIs) were adopted as the effect size for continuous outcomes. Forest plots were generated to visually assess the effect size, corresponding 95% CIs, and heterogeneity. Heterogeneity was assessed by conducting the chi-squared test and I². The values of I² <25%, 25–50%, and >50% indicated low, moderate, and high heterogeneity, respectively. A random model was used if there was significant heterogeneity (I² >50%); otherwise, a fixed model was used. Subgroup analyses were based on the type of acupuncture (electro-acupuncture or manual acupuncture), treatment duration (short-term: ≤ 4 weeks or long-term: >4 weeks), and the type of medication. Sensitivity analyses were performed by limiting our analyses to high-quality studies to determine the robustness of the meta-analysis. Funnel plots were utilized, and Begg's test was conducted to detect the publication bias when at least 10 studies were reporting the outcome(s). The statistical analysis was performed by a statistician (LHJ) using R software, version 4.3.2.

A total of 13,831 records were identified from eight databases (Figure 1). After excluding duplicate records, 10,764 records were screened. A total of 2,168 full-text records were assessed for eligibility through preliminary screening. Finally, 110 studies met the inclusion and exclusion criteria and were included in our review and quantitative synthesis.

Out of 110 RCTs included in our study, the sample sizes ranged from 30 to 440. A total of 11 studies specifically concerned fatigue, 32 concerned depression or anxiety, 7 concerned mild cognitive impairment, 18 concerned headache, and 42 concerned insomnia. The overall treatment duration varied from 10 days to 4 months. Most of the trials were two-armed, with the exception of one four-armed trial and 14 three-armed trials. The detailed characteristics of each study are presented in Supplementary Table 3.

Figure 2 shows the overall risk of bias, and the details of each study are shown in Supplementary Figure 1. A total of 79 out of 110 studies reported the randomization process. Approximately 84% of the 79 articles specified the generation of random sequences and 82% stated the allocation concealment. A total of 31 studies were rated as having a high risk of performance bias or unclear performance bias (blinding of participants and personnel). Out of these 79 studies, 64 performed single blinding of patients by using true and sham techniques for stimulating acupoints, and 59 studies reported using blinded assessors. No study was double-blinded as it is difficult to blind the acupuncture therapist. Other biases were caused by unreported dropouts or a large number of dropouts affecting the results (22–27), a change in protocol during the course of the trial (22, 28–35), potential carryover effects in crossover studies (23, 25, 26, 28, 33, 35–50), and problems associated with the statistical analysis of results and the tendency to draw optimistic conclusions (25, 26, 28, 35, 36, 38, 42, 43, 47, 51–55). In conclusion, the assessment of the risk of bias highlights crucial methodological considerations. These revelations necessitate meticulous consideration when interpreting the outcomes of the meta-analysis.

A total of 110 studies with 8,514 participants were included in the meta-analysis. The results of the meta-analysis are presented in Table 1. The forest plots of the meta-analysis are shown in Figures 3–7.

Outcome 1. Fatigue Scale (FS-14)

Comparison 1. Acupuncture vs. medication

One study compared acupuncture and medication for the treatment of fatigue as measured using the FS-14 (0–14 scores) and found that acupuncture could relieve the symptoms of fatigue (MD: −2.27, 95% CIs: −2.85 to −1.69, P < 0.01) (Figure 3) (56).

Comparison 2. Acupuncture vs. sham acupuncture

Five trials that compared acupuncture with sham acupuncture were included (57–61). The result indicated that acupuncture could reduce the scores of the FS−14 and improve fatigue symptoms compared to sham acupuncture (MD: −3.36, 95% CIs: −4.91 to −1.79, P < 0.01), although heterogeneity was present (I² = 95%).

Outcome 2. Fatigue Assessment Instrument (FAI)

Two studies compared acupuncture with moxibustion (62, 63). The pooled analysis indicated that acupuncture is inferior to moxibustion in treating fatigue symptoms (MD: 27.46, 95% CIs: 20.17 to 34.75, I² = 0%, P < 0.01).

Outcome 3. Visual Analog Scale (VAS)

Four studies compared acupuncture with sham acupuncture (59, 64–66), and the results showed that acupuncture could decrease chronic fatigue–related pain (MD: −36.02, 95% CIs: −63.30 to −8.75, I² = 88%, P = 0.01).

Outcome 1. Hamilton Depression Rating Scale (HAMD)

Comparison 1. Acupuncture vs. medication

Twelve studies involving 738 participants were included in this pooled analysis (28, 36–38, 51, 52, 67–72). The results showed that acupuncture was better than medication in improving depressive symptoms (MD: −1.62, 95% CIs: −3.07 to −0.17, P < 0.01). However, there was substantial heterogeneity among these studies (I² = 85%) (Figure 4).

The subgroup analysis based on treatment duration (≤ 4 weeks vs. >4 weeks) indicated that short–term efficacy of acupuncture was better (≤ 4 weeks: MD: −2.96, 95% CIs: −5.47 to −0.45; >4 weeks: MD: −0.68, 95% CIs: −2.40 to 1.03) than medication. The subgroup analysis of electro–acupuncture or manual acupuncture failed to show significant benefits for treating depression compared with medication (electro–acupuncture: MD: −0.89, 95% CIs: −1.82 to 0.05; manual acupuncture: MD: −1.80, 95% CIs: −3.84 to 0.24) (Supplementary Figure 2). However, there was heterogeneity among the subgroups of ≤ 4 weeks, >4 weeks, and manual acupuncture.

A sensitivity analysis was conducted on the remaining high–quality studies assessed by the RoB 2 Development Group to investigate the influence of study quality on the effect size. The result of the sensitivity analysis showed that acupuncture was not superior to medication (fluoxetine) in treating depression (MD: −1.32, 95% CIs: −6.23, 3.59; P <0.01). However, the heterogeneity was still significant (I² = 93%) (Supplementary Figure 3).

Comparison 2. Acupuncture vs. sham acupuncture

In four studies (73–76), acupuncture was shown to have better effectiveness than sham acupuncture in relieving depressive symptoms (MD: −9.47, 95% CIs: −13.74 to −5.19, I²=95%, P < 0.01).

Comparison 3. Acupuncture vs. waitlist

One study reported that acupuncture was more effective in relieving depressive symptoms than the control waitlist group (MD: −3.80, 95% CIs: −6.96 to −0.64, P = 0.02) (77).

Comparison 4. Acupuncture plus medication vs. medication

Two studies reported that acupuncture with medication was more effective in relieving depressive symptoms than medication alone (MD: −4.87, 95% CIs: −9.53 to −0.21, I² = 35%, P < 0.01) (78, 79).

Comparison 5. Acupuncture plus rTMS vs. sham acupuncture plus rTMS

One study compared acupuncture plus Repetitive Transcranial Magnetic Stimulation (rTMS) with sham acupuncture plus rTMS (80). The acupuncture plus rTMS group demonstrated better efficacy (MD: −4.28, 95% CIs: −6.63, −1.93, P < 0.01).

Outcome 2. Hamilton Anxiety Rating Scale (HAM–A)

Nine studies compared acupuncture with medication using the HAM–A scores (30–35, 39, 81, 82). These studies did not show any significant difference between the two groups (MD: −0.65, 95% CIs: −1.65, 0.35, I² = 83%, P < 0.01).

Two studies compared acupuncture with sham acupuncture and did not show any significant difference between the two groups (MD: −7.78, 95% CIs: −17.06, 1.51, I² = 93%, P = 0.10) (73, 75).

One study compared acupuncture with Cranial Electrotherapy Stimulation (CES) (82). There was no significant difference between the two groups (MD: −0.41, 95% CIs: −1.02, 0.20, P = 0.19).

Outcome 3. Self–Rating Depression Scale (SDS)

Four studies compared acupuncture with medication using the SDS and found no significant difference between the two groups (MD: −0.45, 95% CIs: −6.30, 5.41, I² = 86%, P = 0.88) (22, 29, 38, 70).

Four studies compared acupuncture with sham acupuncture, and it was found that acupuncture could reduce the scores of the SDS (MD: −13.57, 95% CIs: −19.24, −7.90, I² = 90%, P < 0.01) (29, 74, 83, 84).

Outcome 1. Mini-Mental State Examination (MMSE)

Comparison 1. Acupuncture vs. medication

Five studies compared acupuncture with medication using the MMSE (ranging from 0 to 30) (23, 85–88) and showed that acupuncture was slightly better than medication in improving mild cognitive impairment (MD: 1.15, 95% CIs: 0.88 to 1.41, P < 0.01), without significant heterogeneity (I² = 0%) (Figure 5).

Comparison 2. Acupuncture vs. sham acupuncture

A pooled analysis of two studies indicated that acupuncture could slightly relieve the symptoms of mild cognitive impairment compared with sham acupuncture (MD: 1.20, 95% CIs: 0.50 to 1.90, I²=0%, P <0.01) (53, 89).

Comparison 3. Acupuncture vs. waitlist

One study indicated that acupuncture could relieve cognitive impairment symptoms compared with the waitlist (MD: 2.38, 95% CIs: 1.43 to 3.33, P < 0.01) (89).

Outcome 1. Headache diary records (headache frequency per month)

Two studies compared the overall scores between acupuncture and medication (24, 90), and no significant difference was reported (MD: 0.03, 95% CIs: −0.86 to 0.91, I² = 0%, P = 0.95) (Figure 6).

Four studies compared acupuncture with sham acupuncture (54, 91–93), and no obvious difference was found between the two groups (MD: −3.35, 95% CIs: −7.13 to 0.44, I² = 83%, P = 0.08).

Two studies compared acupuncture with the waitlist control (93, 94), and no significant difference was detected (MD: −3.77, 95% CIs: −8.67 to 1.12, I² = 93%, P = 0.13).

One study compared acupuncture with Transcutaneous Electrical Nerve Stimulation (TENS) (95), and it showed that TENS was better than acupuncture for relieving headache (MD: 6.61, 95% CIs: 3.66 to 9.56, P < 0.01).

Outcome 2. VAS

Comparison 1. Acupuncture vs. medication

Two studies compared acupuncture with medication using the VAS scores for headache. The results showed that acupuncture is slightly better than medication therapy in reducing headache (MD: −1.05, 95% CIs: −1.48 to −0.62, I² = 0%, P <0.01) (55, 96).

Comparison 2. Acupuncture vs. sham acupuncture

Six studies compared acupuncture with sham acupuncture (40, 54, 97–100). The results showed that acupuncture was not superior to sham acupuncture for relieving headache (MD: −0.74, 95% CIs: −1.70 to 0.23, I² = 84%, P = 0.13).

Comparison 3. Acupuncture vs. waitlist

One study compared acupuncture with waitlist control (101). It showed a slight difference between the two groups (MD: −0.48, 95% CIs: −0.93 to −0.03, P = 0.04).

Outcome 3. Migraine Disability Assessment Scale (MIDAS)

One study compared acupuncture with medication using the MIDAS and showed that acupuncture could reduce the overall scores after treatment compared with medication (MD: −10.30, 95% CIs: −16.82 to −3.78, P <0.01) (102).

One study compared acupuncture with sham acupuncture (103). There was no significant difference reported between the two groups (MD: −9.00, 95% CIs: −20.86 to 2.86, P = 0.14).

One study compared acupuncture with the waitlist (101). It failed to show any significant difference between the two groups (MD: −9.60, 95% CIs: −24.69 to 5.49, P = 0.21).

Outcome 1. Pittsburgh Sleep Quality Index (PSQI)

Comparison 1. Acupuncture vs. medication

A total of 18 studies were included in the analysis of insomnia measured using the PSQI (0–21 scores) (25, 26, 41–46, 48, 104–112). Acupuncture was reported to have better efficacy for insomnia than medication (MD: −2.33, 95% CIs: −2.99 to −1.67, I² = 87%, P < 0.01) (Figure 7).

The subgroup analysis showed that there was a slight difference between the electro–acupuncture and manual acupuncture groups (electro–acupuncture: MD: −2.71, 95% CIs: −3.81 to −1.61; manual acupuncture: MD: −2.31, 95% CIs: −3.00 to −1.61; P =0.54) (Supplementary Figure 2). In addition, alprazolam showed better efficacy than other medications. A second subgroup analysis was not conducted in the present study because the treatment duration of all included studies was <4 weeks.

A sensitivity analysis was carried out exclusively on studies with a low risk of bias. Among six high–quality studies, it was shown that acupuncture provided a slight benefit for treating insomnia compared with medication (estazolam) (MD: −3.44, 95% CI: −4.39, −2.50, P < 0.01). This result was consistent with the findings of the total meta–analysis (Supplementary Figure 3).

Comparison 2. Acupuncture vs. sham acupuncture

A total of 21 studies compared the efficacy of acupuncture with that of sham acupuncture (27, 60, 76, 84, 113–129). The overall meta–analysis result showed that acupuncture had better efficacy for treating insomnia (MD: −4.19, 95% CIs: −5.33 to −3.05, I²=93%, P < 0.01). The results of the two subgroup analyses indicated that manual acupuncture had better efficacy than electro–acupuncture, and acupuncture was found to be more effective in the short term than in the long term (Supplementary Figure 2).

The sensitivity analyses solely included high–quality studies, which showed that acupuncture was beneficial for treating insomnia compared with sham acupuncture (MD: −4.36, 95% CI: −5.22, −3.50, P < 0.01). The finding was consistent with the result of the meta–analysis (Supplementary Figure 3).

Comparison 3. Acupuncture vs. non–invasive intervention

One study compared acupuncture with acupressure (49). There was no difference between the two groups (MD: 0.74, 95% CIs: −0.63 to 2.11, P = 0.29).

Comparison 4. Acupuncture plus rTMS vs. sham acupuncture plus rTMS

One study compared acupuncture plus rTMS with sham acupuncture plus rTMS (80). The acupuncture plus rTMS group showed better efficacy (MD: −2.38, 95% CIs: −4.40, −0.36, P = 0.02).

Outcome 2. Insomnia Severity Index (ISI)

One study compared acupuncture with medication using the ISI and showed that acupuncture could reduce the overall scores after treatment compared with medication (MD: −2.20, 95% CIs: −3.71 to −0.69, P <0.01) (130).

Four studies compared acupuncture with sham acupuncture. The acupuncture group did not show better efficacy (MD: −5.19, 95% CIs: −10.46 to 0.09, I² = 97%, P = 0.05) (83, 122, 130, 131).

Outcome 3. Score changes measured using the Epworth Sleepiness Scale (ESS)

Two studies compared acupuncture with medication using the ESS and showed that acupuncture could reduce the overall scores compared with medication (MD: −4.64, 95% CIs: −5.80 to −3.48, I² = 0%, P < 0.01) (47, 50).

A total of 45 out of 110 (40.91%) included studies reported that patients with adverse effects were observed in both acupuncture and control groups (Supplementary Table 3). Among the 2,917 patients included in the 45 studies, 227 (7.78%) experienced adverse effects during acupuncture treatment, including instances of local bleeding, hematoma, needling pain, and dizziness. These effects were generally mild in severity, and the patients recovered well after rest or simple treatment. A total of 115 patients out of 1,153 (9.97%) taking Western medication reported adverse effects. It was reported that the incidence of acupuncture-related adverse effects was not significantly higher than that of medication treatment. A total of 14 patients out of 2,917 (0.48%) receiving acupuncture treatment withdrew from the clinical trial due to intolerance to acupuncture or adverse effects. A total of 9 patients out of 2,917 (0.31%) receiving acupuncture treatment reported severe adverse effects, but all of these effects were found to not be due to the intervention.

The Begg's tests did not show any significant difference, and the p-values were 0.6209 (acupuncture vs. medication for insomnia), 0.3021 (acupuncture vs. sham acupuncture for insomnia), and 0.6810 (acupuncture with medication for depression). The visual inspection of the funnel plots also did not reveal any substantial asymmetry in these comparisons (Supplementary Figure 4).