생리통

Background: Dysmenorrhea has a significant negative impact on teenagers’ quality of life, and its prevalence is increasing annually. Although studies have explored the factors affecting dysmenorrhea, it remains unclear how these factors interact with one another. This study aimed to explore the mediating role of binge eating and sleep quality between depression and dysmenorrhea.

Methods: This cross-sectional study recruited adolescent girls from the Health Status Survey of adolescents in Jinan, Shandong Province, and used multistage stratified cluster random sampling. Data was collected using an electronic questionnaire between March 9, 2022, and June 20, 2022. The Numerical Rating Scale and Cox Menstrual Symptom Scale were used to assess dysmenorrhea and the Patient Health Questionnaire-9 to assess depression. The mediation model was tested by Mplus 8.0, and the mediating effect was analyzed using the Product of Coefficients approach and the Bootstrap method.

Results: Among the total of 7818 adolescent girls included in this study, the prevalence of dysmenorrhea is 60.5%. A significant positive association was found between dysmenorrhea and depression. Binge eating and sleep quality seemingly mediate this association. The mediating effect of sleep quality (21.31%) was greater than that of binge eating (6.18%).

Conclusions: The findings of this study point in the right direction for preventing and treating dysmenorrhea in adolescents. For adolescent dysmenorrhea, mental health should be considered and proactive steps taken for educating adolescents on healthy lifestyles to reduce negative consequences of dysmenorrhea. Longitudinal studies on the causal link and influence mechanisms between depression and dysmenorrhea should be conducted in the future.

Supplementary Information: The online version contains supplementary material available at 10.1186/s12905-023-02283-6.

Sample characteristics: The mean age of the 7,818 adolescent girls included in this study was 15.71 ± 1.617 years. The prevalence of dysmenorrhea is 60.5% (n = 4732), and that of mild, moderate, severe, and very severe dysmenorrhea was 12.8%, 33.7%, 12.4%, and 1.6%, respectively. In addition, information about the frequency of CMSS and PHQ-9 are shown in Supplementary Table 3 and 4.

Table 1 summarizes the sociodemographic information of the 4,732 girls with dysmenorrhea and the univariate analysis of the total CMSS scale score in the subgroups of different category variables. Univariate analysis showed that age, type of school, boarding at school, maternal history of dysmenorrhea, sleep duration, sleep quality, and binge eating were significantly associated with dysmenorrhea.Table 1Characteristics of the sample (N = 4732)VariablesN(%)CMSS ScoreH/ZPM(P25-P75)Observation variable4732(100.0)27.0(14.00–45.00)Age59.259a0.000 10–13394(8.3)19.0(8.75–34.00) 14–173611(76.3)27.0(14.00–46.00) 18–20727(15.4)30.0(15.00–48.00)Type of school173.956a0.000 Junior high school1776(37.5)21.0(10.00–37.00) Vocational high school844(17.8)33.0(18.00–56.00) Senior high school2112(44.6)30.0(15.25–48.00)Father's education0.586a0.746 Junior high school and below2371(50.1)27.0(14.00–45.00) Senior high school1242(26.2)26.0(13.00–45.00) Bachelor’s/college degree and above1119(23.6)27.0(14.00–45.00)Mother’s education3.582a0.167 Junior high school and below2677(56.6)26.0(13.00–45.00) Senior high school1101(23.2)26.0(13.00–45.00) Bachelor’s/college degree and above954(20.2)29.0(14.00–46.00)Boarding at school2.198b0.028 Yes2130(45.0)28.0(14.00–47.00) No2602(55.0)26.0(14.00–44.00)Mother's history of dysmenorrhea49.071a0.000 Yes2305(48.7)29.0(16.00–48.00) No1050(22.2)23.0(10.00–41.00) No clear1377(29.1)27.0(14.00–45.00)Sleep duration61.692a0.000 Inadequate3348(70.8)29.0(15.00–47.00) Normal1351(28.6)23.0(10.00–40.00) Excessive33(0.7)30.0(21.00–59.50)Sleep quality906.332a0.000 Good1953(41.3)17.0(8.00–32.00) General2195(46.4)31.0(18.00–46.00) Poor584(12.3)55.0(36.00–76.00)Binge eating265.551a0.000 Often126(2.7)43.0(19.75–69.00) Sometimes1223(25.8)36.0(21.00–56.00) Never3383(71.5)23.0(11.00–40.00)CMSS the Cox Menstrual Symptom ScaleaKruskal-wallis testbMann-Whitney U test

Preliminary analyses: Table 2 presents the Spearman’s rank correlation coefficients for the main study variables. Depression was positively associated with dysmenorrhea (r = 0.543, p < 0.01), binge eating behavior (r = 0.262, p < 0.01), and poor sleep quality (r = 0.334, p < 0.01). Binge eating behavior was positively associated with dysmenorrhea (r = 0.237, p < 0.01) and poor sleep quality (r = 0.112, p < 0.01). Poor sleep quality was positively associated with dysmenorrhea (r = 0.428, p < 0.01).Table 2The correlations among the main study variablesVariables1234Depression score1CMSS score0.543**1Binge eating0.262**0.237**1Sleep quality0.334**0.428**0.112**1CMSS the Cox Menstrual Symptom Scale** p < 0.01

Testing for the mediating effect: The mediator model showed an acceptable fit: χ2/df = 1.705, CFI = 0.994, TLI = 0.990, RMSEA = 0.012, and SRMR = 0.059.

As shown in Table 3 and Fig. 1, binge eating and sleep quality had partial mediating effects on depression and dysmenorrhea, respectively. The difference in the mediating effects between binge eating and sleep quality was statistically significant (Table 4), and the mediating effect of sleep quality (21.31%) was greater than that of binge eating (6.18%); however, the binge eating to sleep quality pathway was not statistically significant (P = 0.213).Table 3The chained mediation model and 95% CIsModel pathwaysPointEstimateProduct of CoefficientsBootstrap 5000 Times95%CIS.EEst./S.EP ValueLowerUpperPath A0.1070.00813.6440.0000.0920.122Path B0.0310.0056.2170.0000.0220.042Path C0.0020.0021.2450.213-0.0020.006Path D0.3620.01820.4750.0000.3230.391Total indirect0.1400.00915.5180.0000.1230.158Total0.5020.01534.5420.0000.4670.565Adjusted for age, school type, educational level of parents, boarding at school, mother's history of dysmenorrhea, and sleep duration. The values listed in the table are all standardized values. S.E Standard errorPath A: Depression- > Sleep Quality- > DysmenorrheaPath B: Depression- > Binge Eating- > DysmenorrheaPath C: Depression- > Binge Eating- > Sleep quality- > DysmenorrheaPath D: Depression- > DysmenorrheaFig. 1Mediating effects of binge eating and sleep quality on depression and dysmenorrhea. Note: adjusted for age, school type, educational level of parents, boarding at school, mother's history of dysmenorrhea, and sleep duration. Standardized pathway coefficients are shown outside in parentheses, and standard errors are shown in parentheses, * p < 0.01Table 4The parallel mediation, 95% CIs and comparison between the mediation effectsModel pathwaysPointEstimateProduct of CoefficientsBootstrap 5000 Times95%CIS.EEst./S.EP ValueLowerUpperIndirect Effects Path AB10.1530.0236.6330.0000.1120.205 Path AB20.5000.03514.2050.0000.4360.575 Total indirect0.6530.04315.0630.0000.5780.747 Direct1.6260.08020.2250.0001.4641.775 Total2.2790.04732.1350.0002.1372.418Contrast Path AB1 vs. Path AB2-0.3470.041-8.4910.000-0.428-0.270Note: adjusted for age, school type, educational level of parents, boarding at school, mother's history of dysmenorrhea, and sleep duration. The values listed in the table are all non-standardized values. S.E Standard errorPath AB1: Depression- > Binge eating- > DysmenorrheaPath AB2: Depression- > Sleep Quality- > Dysmenorrhea

Background: Sleep deprivation is known to be a risk factor for premenstrual syndrome and primary dysmenorrhea in adults. However, it has rarely been investigated in adolescents. The aim of this study was to investigate whether sleep pattern, duration, and quality independently affect premenstrual syndrome and primary dysmenorrhea in adolescent girls. An additional purpose was to investigate the sleep status in Korean adolescent girls during the COVID-19 pandemic.

Methods: A cross-sectional survey study was conducted in 519 high school girls aged 15 to 18 years in Gyeonggido, South Korea, in 2021 during the COVID-19 lockdown. Menstrual pain intensity and menstrual symptoms were assessed using the visual analogue scale (VAS) and Cox menstrual symptom scale (CMSS), respectively. Premenstrual syndrome was assessed by the premenstrual symptoms screening tool (PSST). Sleep was assessed by the Pittsburgh Sleep Quality Index (PSQI). The known risk factors of dysmenorrhea, including menstrual and lifestyle characteristics and stress, were assessed as covariates.

Results: During the pandemic, approximately 68% of girls slept 7 h or less, while about 60% reported poor sleep quality. Additionally, 64% of participants had a bedtime later after 1AM, and 34% woke up later after 8AM. Late bedtime significantly affected VAS (P = 0.05), CMSS severity and frequency (both P < 0.01), and PSST symptom (P < 0.01). Waking up late affected CMSS severity (P < 0.05), PSST symptom (P = 0.05), and PSST function (P < 0.05). However, the significance of these effects disappeared after controlling for covariates. Sleeping less than 5 h affected CMSS frequency (P < 0.05) and PSST symptoms (P < 0.001). After controlling for covariates, the significance of the effect on PSST symptom remained (P < 0.05). General sleep quality and PSQI components, including subjective sleep quality, sleep latency, sleep disturbance, use of sleeping medication, and daytime dysfunction, significantly affected CMSS frequency and severity and PSST symptom after controlling for covariates (P < 0.05, P < 0.01, or P < 0.001). The multiple regression analysis revealed that among sleep characteristics, sleep quality was the most important risk factor of premenstrual syndrome and dysmenorrhea.

Conclusion: Our study result heightens the importance of healthy sleep hygiene, especially sleep quality in the management of premenstrual syndrome and dysmenorrhea in adolescent girls.

Participant characteristics: Most participants had dysmenorrhea, with 17.3% (n = 90) experiencing mild pain (0 < VAS < 3) and 76.5% (n = 397) experiencing moderate-to-severe pain (3 ≦ VAS), while 6.2% had no pain (VAS = 0) when dysmenorrhea was defined by the intensity of menstrual pain measured by VAS. Regarding classification of PMS defined by PSST, 60.9% of participants (n = 316) reported either no symptoms or mild symptoms, whereas 39.1% of them (n = 203) experienced moderate-to-severe PMS symptoms. Additionally, 4.8% of the participants (n = 25) did not report any of the 14 PMS symptoms and 5 functional impairments.

Characteristics of 519 study participants are presented in Table 1. The study participants were high-school students (1st year to 3rd year) with a mean age of 16.2 ± 0.8 years (aged 15 to 18 years). Most participants had their menarche after 12 years (71.7%) and had a normal length of menstrual cycle (89.2%) and period length (86.7%). About half of them had an irregular menstrual cycle, and 22.5% of participants had heavy menstrual volume. About 36% of them had a family history of dysmenorrhea. The majority of participants were non-smokers (94.4%), did not drink alcohol beverages (79.8%), ate breakfast (77.8%), consumed caffeine beverages (60.9%), and did not engage in moderate or vigorous exercise (79.4%). The average PSS score of participants was 20.6 ± 6.1, and the mean sleep duration of participants was 6.2 ± 1.6 h. Overall, 59.7% of the total sample had a global PSQI score of more than 5, indicating poor sleep quality.

Differences in participant characteristics between the groups with good and poor sleep quality are presented in Table 1. Among menstrual and lifestyle characteristics, period regularity (P < 0.001), volume of menstrual fluid (P < 0.01) and caffeine consumption (P < 0.01) showed the significant group-differences, revealing that proportions of irregular menstrual cycle, heavy menstrual volume, and caffeine consumer were higher in the group with poor sleep quality. There were no group-differences in all other menstrual and lifestyle characteristics, age, and BMI. Significant group-differences were observed in PSS, sleep duration, and the global PSQI score, as well as average scores for the 7 PSQI components (P < 0.01 or P < 0.001), revealing a higher stress level, higher PSQI scores, and shorter sleep time in the group with poor sleep quality.

Sleep and menstrual characteristics during the COVID-19 pandemic: The sleep pattern of Korean high school girls during the COVID-19 pandemic and menstrual differences according to sleep pattern are shown in Table 2. The proportion of participants who went to bed between 10PM and 1AM was 36%, and the rest of them slept after 1AM. The CMSS severity and frequency, as well as the number of moderate-to-severe PSST symptoms, were higher in girls who slept after 1AM, compared to those who slept earlier (all P < 0.01). Most girls got up between 5AM and 8AM (66.5%),while 33.5% of them woke up later.

Girls who got up late had worse CMSS severity (P < 0.05) and PSST functional impairments (P < 0.05). Approximately 60% of high school girls had a poor sleep quality. The VAS, CMSS severity and frequency, and PSST symptoms and functional impairment were worse in girls with poor sleep quality compared to those with good sleep quality (P < 0.01 or P < 0.001). Half of the girls (53.6%) slept 5–7 h, 32.4% of them slept more than 7 h, and 14.1% slept less than 5 h. The CMSS frequency tended to be higher in girls who slept less than 5 h compared to those who slept more than 7 h (P = 0.06). Additionally, the number of moderate-to-severe PSST symptoms was higher in girls who slept less than 5 h, compared to those who slept 5–7 h or more than 7 h (both P < 0.01).

Correlations between sleep and menstrual variables: The results of Pearson correlation coefficients are presented in Table 3. Sleep duration was not associated with any of the five menstrual variables related to dysmenorrhea and premenstrual symptoms. The correlations between the global PSQI score and all menstrual variables (VAS, CMSS severity and frequency, and number of PSST symptoms and functional impairment) were found to be significant (all P < 0.001), indicating that poorer sleep quality was associated with more severe dysmenorrhea and premenstrual symptoms. All five menstrual variables were positively correlated with five of the seven PSQI component scores: subjective sleep quality, sleep latency, sleep disturbance, use of sleeping medication, and daytime dysfunction (P < 0.05, P < 0.01 or P < 0.001), showing that better scores on these components were associated with less severe dysmenorrhea and premenstrual symptoms. The PSQI sleep duration was associated with the number of moderate-to-severe PSST symptoms (P < 0.05).

Impacts of sleep variables on menstrual variables: The result of significant unstandardized and standardized regression coefficients from simple and multiple regression analyses are presented in Table 4 (Vertical line of the first row should be adjusted as shown in Table 5.). The results of simple regression analyses regarding effect of bedtime (10PM-1AM vs. After 1AM) on variables of dysmenorrhea and premenstrual symptoms showed that bedtime affected VAS (P = 0.05), the CMSS severity and frequency (both P < 0.01), and the number of PSST symptoms (P < 0.01). These findings indicate that going to bed late after 1AM increased menstrual pain intensity, the severity and frequency of menstrual symptoms, and the number of moderate-to-severe premenstrual symptoms. However, the results of multiple linear regression analyses showed that the significance of the effect of bedtime disappeared after controlling for other risk factors for dysmenorrhea (covariates).

The unadjusted β of wake-up time was significant for CMSS severity (P < 0.05) and the number of PSST symptoms (P = 0.05) and functional impairment (P < 0.05), indicating that getting up between 5AM and 8AM was associated with less severe menstrual symptoms and a lower occurrence of moderate-to-severe premenstrual symptoms and functional impairment, compared to getting up later in the morning and in the afternoon. However, when covariates were entered, the adjusted β was not significant.

The unadjusted β of sleep duration (< 5 h vs. ≥5 h) for CMSS frequency (P < 0.05) and PSST symptoms (P < 0.001) were significant, indicating that sleeping less than 5 h was associated with more frequent occurrence of menstrual symptoms and worse premenstrual symptoms. After controlling for covariates, only the adjusted β for PSST symptoms remained significant (P < 0.05).

The results of simple regression analyses showed that general PSQI sleep quality (poor vs. good) had significant effects on all 5 dependent variables, including VAS, CMSS severity and frequency, and the number of PSST symptoms and functional impairment (all P < 0.001). These results indicated that girls with poor sleep quality experienced more severe menstrual pain and symptoms, more frequent menstrual symptoms, and more severe premenstrual symptoms and functional impairments. In multiple regression analyses, after controlling for covariates, the effect of sleep quality remained significant for the CMSS severity (P < 0.01) and frequency (P < 0.001), as well as the number of moderate-to-severe PSST symptoms (P < 0.01).

Among the 7 PSQI components, 5 components (subjective sleep quality, sleep latency, sleep disturbance, daytime dysfunction, and use of sleep medication) significantly affected all 5 dependent variables (P < 0.05, P < 0.01, or P < 0.001), showing that higher components scores resulted in worse dysmenorrhea and premenstrual symptoms. After controlling for covariates, all 5 PSQI components significantly affected the CMSS frequency and severity and PSST symptoms and functional impairments (P < 0.05, P < 0.01, or P < 0.001), while only sleep disturbance significantly affected VAS (P < 0.01). The PSQI sleep duration and habitual sleep efficiency were not significant factors affecting dysmenorrhea and premenstrual symptoms.

The most important sleep-related risk factors of dysmenorrhea and premenstrual symptoms: To examine which variables were the most important risk factors of dysmenorrhea and premenstrual symptom among sleep variables, multiple regression analyses were performed. Bedtime, wake-up time, sleep duration, and general sleep quality were entered together in the regression model with all covariates. As shown in Table 5, the results of multiple regression analyses showed that the significant risk factors of VAS included PSS (P < 0.001) and menstrual volume (P < 0.001). The significance of the adjusted beta coefficient for sleep quality was close to the significance level (P = 0.058). The significant risk factors of the CMSS severity included general sleep quality (P < 0.001), PSS (P < 0.001), menstrual volume (P < 0.001), family history of dysmenorrhea (P < 0.001), having menarche before 12 years (P < 0.05), eating breakfast (P < 0.01), and drinking (P < 0.01). The significant risk factors of the CMSS frequency included general sleep quality (P < 0.001), PSS (P < 0.001), menstrual volume (P < 0.001), family history of dysmenorrhea (P < 0.01), and eating breakfast (P < 0.01).

The significant risk factors of the PSST symptoms included general sleep quality (P < 0.01), PSS (P < 0.001), menstrual volume (P < 0.01), having menarche before 12 years (P < 0.05), and eating breakfast (P < 0.05). In addition, the significant risk factors for PSST functional impairment included PSS (P < 0.001) and menstrual volume (P < 0.01). The significance of the adjusted beta coefficients for general sleep quality (P = 0.054) and drinking (P = 0.054) was close to the significance level.

Supplemental Digital Content is available in the text

Study selection: A total of 4244 articles were screened, and 3962 were retrieved. The full texts of 282 studies were reviewed; 222 did not meet our inclusion criteria. Finally, 60 RCTs meeting our criteria were included. All studies were published between January 1987 and November 2017. Forty-four studies were published in Chinese, 15 in English, and 1 in Korean.Figure 1 shows a Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) flow of the study selection process.

Study characteristics / Patients: A total of 3171 patients were treated with MA, EA, warm acupuncture (WA), AA, or catgut embedding therapy (CET); 2730 control patients received no treatment, placebo acupuncture, or oral medications. Of 60 trials, 55 were conducted in China (5653 patients), and 1 each was conducted in America (22 patients), Turkey (35 patients), Australia (92 patients), Thailand (52 patients), and South Korea (47 patients), respectively. The age range of the participants was 10 to 43 years. Table 1 summarizes the characteristics of the included studies.

Acupuncture interventions: Of 60 trials, 35 used MA, 11 used EA, 11 used WA, 1 used AA, and 2 used CET. The number of acupoints used varied from 1 to 21. The most frequently used point was Sanyinjiao (SP6), followed by Guanyuan (CV4), Diji (SP8), Cialiao (BL32), Zusanli (ST36), Xuehai (SP10), Taichong (LR3), Zhongji (CV3), Shiqizhui (EX-B8), and Shenshu (BL23). The only point used in 9 trials that used EA was Sanyinjiao (SP6). Twenty-one trials used different acupoints or added acupoints based on traditional Chinese medicine patterns. Treatment duration ranged from one day to 3 menstrual cycles; 25 trials included follow-ups, which varied from 180 minutes to one year. The time of intervention started before menstruation started in 31 trials, when menstruation started in 4 trials, when pain occurred in 10 trials, and continuous treatment except for menstrual periods in 6 trials.De-qi sensation was performed in most trials, but 4 studies did not mention about De-qi sensation. Additional interventions to acupuncture were included in 15 trials.Table 2 shows the acupuncture points and treatment methods of the included studies based on STandards for Reporting Interventions in Clinical Trials of Acupuncture (STRICTA) recommendations.

Control interventions: Of the 35 trials that used MA, 14 compared MA to no treatment, 3 compared MA to placebo acupuncture, and 18 compared MA to oral medications. Most of the medications were NSAIDs, and only 1 was an OC. Of the 11 trials that used EA, 4 compared EA to nonacupoint EA (placebo EA), or no treatment, 5 compared EA to nonacupoint EA, and 2 compared EA to NSAIDs. Of 11 trials that used WA, 2 trials compared WA plus NSAIDs to NSAIDs and 9 trials compared WA to NSAIDs. One trial compared AA to NSAIDs. Two trials compared CET to NSAIDs. All of the placebo controls used nonacupoint acupuncture, not sham acupuncture.

Outcome measures: Twenty-seven trials measured pain intensity using VAS, 1 used NRS, 9 used CMSS, 5 used VRS, and 8 used RSS. Thirty trials measured pain relief. Six trials measured overall menstruation symptoms using MSS. Three trials measured the quality of life using SF-36. Twelve trials reported AEs. Finally, of 25 trials conducting follow-ups, 10 trials reported the pain intensity after a follow-up period,,which varied from 180 minutes to 9 months by study. Most studies reported various outcomes measuring dysmenorrhea and related symptoms.

Risk of bias: All 60 studies mentioned randomization. Twenty trials used random number tables, 10 used a computer-generated sequence, 5 used central random method, and 1 used the draw method. Three trials used the order of joining the study, and the other studies did not report the details of randomization. Sixteen studies reported appropriate allocation concealment using computer programs, central telephone controls, sealed envelopes, or independent individual controls.

It is difficult to achieve blinding to both participants and practitioners for the characteristics of study design in acupuncture intervention, but 14 studies mentioned efforts to minimize performance bias, so we assessed the risk of bias as low. Twelve studies reported assessor blinding, but most of the others did not report the details.

Most of the studies had no missing data, performed intention-to-treat (ITT) analysis, or had similar numbers and reasons of drop-outs. However, the details of drop-outs and withdrawals were not reported in 6 studies, considered to be a high risk in reporting bias. Forty-nine studies reported all outcomes clearly as mentioned in protocol studies or methods and were assessed as a low risk of bias in selective reporting. Six studies reported the outcomes unclearly and were assessed as an unclear risk of bias, and 5 studies did not report all outcomes as planned and were assessed as a high risk of bias. There was a low risk of other sources of bias based on lack of clear evidence. As shown in Figures 2 and 3, most of the studies included in this meta-analysis achieved a low or unclear risk of bias of the quality assessment items.

Data synthesis / Manual acupuncture / MA versus no treatment: VAS. Five studies were included in the meta-analysis to synthesize VAS data. As shown in Figure 4A, the pooled results showed serious heterogeneity (I2 = 98%). We conducted a subgroup analysis, and the pooled results showed that after treatment of 1 day, MA more effectively reduced primary dysmenorrhea than no treatment (n = 210, SMD = −1.59, 95% CI [−2.12, −1.06], P <.001, I2 = 60%).

VRS. One study reported that after the treatment of 3 menstrual cycles, MA more effectively reduced primary dysmenorrhea than no treatment (n = 96, MD = −2.04, 95% CI [−2.11, −1.97], P <.001).

CMSS for pain intensity. Six studies were included in the meta-analysis to synthesize CMSS for pain intensity data. As shown in Figure 4B, after treatment of 3 menstrual cycles, MA more effectively reduced pain than no treatment (n = 838, MD = −7.08, 95% CI [−8.53, −5.63], P <.001, I2 = 50%). Two studies reported the subscales of CMSS for pain intensity; 1 study reported the MA significantly reduced abdominal pain, and the other reported that the MA significantly reduced extra bed time.

RSS. Two studies were included for meta-analysis to synthesize RSS data. As shown in Figure 4C, after treatment of 3 menstrual cycles, MA more effectively reduced pain than no treatment (n = 141, MD = −10.47, 95% CI [−10.74, −10.20], P <.001, I2 = 93%).

VAS after follow-up. Five studies were included for meta-analysis to synthesize VAS after follow-up data. As shown in Figure 4D, the pooled results showed serious heterogeneity (I2 = 98%). We conducted a subgroup analysis, and after a 180-minute follow-up, MA was significantly more effective than no treatment (n = 210, SMD = −1.22, 95% CI [−1.53, −0.91], P <.001, I2 = 0%).

MA versus placebo acupuncture: Pain intensity. Three studies reported pain scores, but data were unsuitable for pooling because the score systems of the studies were different from each other. One study reported that MA lowered monthly pain score after treatment of 3 menstrual cycles (n = 22, MD = −70.67, 95% CI [−126.52, −14.82], P =.01) and another study reported that MA lowered the pain score after treatment of 3 menstrual cycles without significant differences (n = 92, MD = −0.7, 95% CI [−1.8, 0.4], P =.21). The other study also reported that there were no significant differences between groups after treatment of 3 menstrual cycles (n = 47).

Pain relief. One study reported that after treatment of 3 menstrual cycles, MA provided a significant improvement in pain compared to placebo acupuncture (n = 22, RR = 2.50, 95% CI [1.12, 5.58], P <.05).

SF-36. One study reported that after treatment of 3 menstrual cycles, there was no significant difference in all SF-36 subscales or both component scores between the 2 groups (n = 92, bodily pain MD = −6.1, 95% CI [−15.1, 2.8], P =.18; General health MD = 5.1, 95% CI [−3.0, 13.2], P =.22; Vitality MD = 3.2, 95% CI [−5.1, 11.6], P =.44; Social function MD = 1.1, 95% CI [−8.0, 10.3], P =.81; Role emotional MD = 2.2, 95% CI [−12.7, 17.1], P =.77; Mental health MD = 6.0, 95% CI [−1.7, 13.6], P =.13; Overall Physical Component MD = −2.3, 95% CI [−5.9, 1.2], P =.19; Overall Mental Component MD = 3.5, 95% CI [−1.1, 8.1], P =.71).

Pain intensity after follow-up. Two studies reported this outcome. One study reported MA maintained pain reduction until 9 months after the completion of treatment (n = 22, MD = −64.90, 95% CI [−122.11, −7.69], P =.03). The other reported there were no significant differences between the groups after 3- and 9-month follow-up periods.

AEs. Two studies reported there were no AEs.

MA versus oral medications: Pain intensity. Five studies comparing MA to NSAIDs reported VAS, and 1 study comparing MA to OCs reported a change in NRS. As shown in Figure 4E, the MA was significantly more effective at reducing pain than NSAIDs (n = 255, SMD = −0.63, 95% CI [−0.88, −0.37], P <.001, I2 = 0%). Meanwhile, OCs were more effective than MA after treatment of 3 menstrual cycles (n = 52, MD = 1.58, 95% CI [0.36, 2.80], P <.01).

Pain relief. Fourteen studies comparing MA to NSAIDs were included for meta-analysis to synthesize TER data. As shown in Figure 4F, MA provided significant pain relief compared to NSAIDs (n = 1,049, RR = 1.17, 95% CI [1.11, 1.22], P <.001, I2 = 0%). The funnel plot of those studies did not show asymmetry. One study reported pain relief as percentage using 6-Likert score, and also showed significant pain relief compared to NSAIDs (n = 84, RR = 2.97, 95% CI [1.75, 5.05], P <.01).

MSS. Three studies comparing MA to NSAIDs reported MSS. As shown in Figure 4G, MA was significantly more effective at improving menstrual symptoms than NSAIDs after treatment of 3 menstrual cycles (n = 190, SMD = −0.53, 95% CI [−0.84, −0.23], P <.001, I2 = 2%).

SF-36. One study reported that there was no significance difference between the 2 groups after treatment of 3 menstrual cycles (n = 52, MD = −1.82, 95% CI [−9.36, 5.72], P =.64).

VAS after follow-up. One study reported that after a 3-month follow-up, MA was significantly more effective than NSAIDs (n = 60, MD = −1.39, 95% CI [−2.65, −0.13], P <.05).

AEs. Two studies reported AEs. One study reported one case of regional discomfort or hemorrhage, 4 cases of headache or myalgia, and 1 case of fever in the MA group, which were all mild. Meanwhile, in the OCs group, 9 cases of abnormal uterine bleeding, 5 cases of headache or myalgia, 3 cases of weight gain, 2 cases of nausea or vomiting, and 1 case of breast bleeding were reported; they were all already known AEs of OCs and were not severe. The other study reported 3 cases of elevated alanine transaminase (ALT), 2 cases of blurred vision, 3 cases of lumbar and leg pain, and 5 cases of others, which were predictable reactions, and soon disappeared.

Electroacupuncture / EA versus no treatment: VAS. Four studies reported that EA was significantly more effective at reducing pain than no treatment (n = 97, MD = −15.56, 95% CI [−22.16, −8.95], P <.001; n = 26, MD = −23.19, 95% CI [−32.06, −14.33], P <.001; n = 20, MD = −22.50, 95% CI [−31.70, −13.30], P <.005; n = 97, P <.001; details of data not shown). Data were unsuitable for pooling for means and SDs were not reported.

VRS. Two studies reported that there was no significant difference between the groups. Data were unsuitable for pooling because they reported the results only in graphs, which made it hard to extract raw data.

RSS. Three studies reported this outcome. Two studies showed there was no significant difference in RSS, but the other study showed that EA was significantly more effective than no treatment in RSS-COX2 (P <.05). Data were unsuitable for pooling because they reported the results only as graphs, which made it hard to extract raw data.

AEs. Four studies reported AEs, but 1study showed 1 case of dizziness after EA.

EA versus placebo acupuncture: VAS. Nine studies reported this outcome; only 6 studies were included in the meta-analysis because the other 3 did not provide SDs. As shown in Figure 5A, the VAS of the EA group was significantly lower than placebo group (n = 826, SMD = −0.32, 95% CI [−0.63, −0.01], P =.04, I2 = 69%). Of the 3 studies excluded from meta-analysis, 1 study reported that EA was significantly more effective in reducing pain than the placebo group in cold-dampness stagnation after one session of treatment (n = 487, MD = −8.2, 95% CI [−13.5, −2.9], P <.005); in other types, there was no significant difference. Another study showed the same result after treatment of 1 menstrual cycle (n = 25, MD = −20.78, 95% CI [−29.82, −11.73], P <.001). The other RCT reported that there was no significant difference between the groups after treatment of 1 menstrual cycle (n = 97, details of data not shown).

VRS. Three studies reported VRS, but only 2 were included in the meta-analysis because the third did not provide SDs. As shown in Figure 5B, after treatment of 1 menstrual cycle, the VRS in the EA group was lower than the placebo group, but there was no significance (n = 347, MD = −0.20, 95% CI [−0.43, 0.03], P =.10, I2 = 61%). The other study also reported a change of VRS in the EA group that was lower than the placebo group, but there was no significance (n = 322, reduction from 3.94 to 3.08 vs reduction 3.72 to 3.02).

RSS. Four studies reported RSS, but the data were unsuitable for pooling due to insufficient reporting. One study reported there was no significant difference between the groups (RSS-COX1: reduction from 18.98 to 18.38 vs reduction from 19.44 to 20.20; RSS-COX2: reduction from 11.88 to 11.38 vs reduction from 10.22 to 10.51), and another study showed no significance after 3 cycles, either (n = 97, RSS-COX1 MD = −0.8, 95% CI [−2.2, 0.7], P =.28; RSS-COX2 MD = −0.5, 95% CI [−1.4, 0.4], P =.28). The other 2 studies also showed no significance in RSS (data not shown).

VAS after follow-up. One study reported that after one-cycle follow-up, there were no significance differences between EA and PA groups (n = 322, MD = −4.40, 95% CI [−9.55, 0.75], P =.09).

AEs. Five studies reported AEs. Three studies reported there were no AEs, and one of the other studies reported one case of dizziness after EA. The other study reported one case of minimal bleeding in the EA group, and one case of minimal bleeding and one case of pain after insertion in the placebo group.

EA versus NSAIDs: VAS. One study reported that after treatment of 3 menstrual cycles, EA was significantly effective at reducing pain than NSAIDs (n = 60, MD = −1.40, 95% CI [−2.21, −0.59], P <.01).

Pain relief. Two studies reported TER, but as shown in Figure 5C, the pooled results showed no significant differences between 2 groups (n = 140, RR = 1.80, 95% CI [0.99, 1.18], P =.09, I2 = 0%).

Auricular acupuncture / AA versus NSAIDs: VAS. One study reported that after treatment of 3 menstrual cycles, there were no significant differences between the 2 groups (n = 70, MD = −0.20, 95% CI [−0.90, −0.50], P =.58).

Warm acupuncture / WA versus NSAIDs: VAS. Three studies were included in the meta-analysis to synthesize VAS data. A meta-analysis of the 3 studies involving 178 participants was implemented, but the results showed serious heterogeneity (I2 = 94%). We conducted a sensitivity analysis by excluding the trial with effect sizes largely different from the others. Statistical heterogeneity was reduced after exclusion. As shown in Figure 6A, with 2 remaining studies, the VAS of the WA group was significantly lower than NSAIDs group (n = 114, SMD = −1.12, 95% CI [−1.81, −0.43], P =.002, I2 = 66%).

TER. Eight studies were included in the meta-analysis to synthesize TER data. A meta-analysis of the 3 studies involving 614 participants was implemented, but the results showed serious heterogeneity (I2 = 77%). We conducted a sensitivity analysis by excluding the trial with effect sizes largely different from the others. As shown in Figure 6B, with 6 remaining studies, the WA provided significant pain relief compared to NSAIDs after the treatment of 3 menstrual cycles (n = 570, RR = 1.22, 95% CI [1.10, 1.35], P <.001, I2 = 53%).

CMSS for pain intensity. One study reported that after the treatment of 3 menstrual cycles, WA more effectively reduced primary dysmenorrhea than NSAIDs (n = 70, MD = −8.00, 95% CI [−9.54, −6.46], P <.001).

MSS. One study reported that there were no significant differences between WA and NSAIDs groups (n = 80, MD = −1.09, 95% CI [−2.65, −0.47], P =.17).

AEs. Two studies reported AEs. One study reported 5 cases of nausea, vomiting, and fever in the NSAIDs group, and the other study reported there were no AEs.

WA plus NSAIDs versus NSAIDs: TER. Two studies reported that WA adding on NSAIDs provided significant pain relief compared to only NSAIDs after treatment of 3 menstrual cycles (n = 160, RR = 1.28, 95% CI [1.12, 1.46], P <.001, I2 = 0%).

Catgut embedding therapy / CET versus NSAIDs: VAS. One study reported that CET was significantly more effective than NSAIDs after treatment of 3 menstrual cycles (n = 70, t = −2.70, P <.01).

TER. Two studies reported that CET provided significant pain relief compared to NSAIDs after treatment of 3 menstrual cycles as shown in Figure 7A (n = 162, RR = 1.40, 95% CI [1.19, 1.65], P <.001, I2 = 54%).

MSS. Two studies reported that CET effectively reduced menstrual symptoms compared to NSAIDs after treatment of 3 menstrual cycles with serious heterogeneity as shown in Figure 7B (n = 162, SMD = −1.57, 95% CI [−1.95, −1.19], P <.001, I2 = 98%).

SF-36. One study reported that SF-36 in the CET group was significantly higher than the NSAIDs group after treatment of 3 menstrual cycles (n = 70, t = 2.535, P <.05).

VAS after follow-up. One study reported that after a 3-month follow-up, CET was significantly more effective than NSAIDs (n = 70, t = −4.72, P <.01).

AEs. One study reported 6 cases of gastrointestinal discomforts, headache, dizziness, and insomnia in the NSAIDs group.

Primary dysmenorrhea, which is menstrual pain without pelvic pathology, is the most common gynecologic condition in women. Heat therapy has been used as a treatment. We assessed the evidence on heat therapy as a treatment for primary dysmenorrhea. We searched 11 databases for studies published through July 2018. All randomized controlled trials (RCTs) that addressed heat therapy for patients with primary dysmenorrhea were included. Data extraction and risk-of-bias assessments were performed by two independent reviewers. Risk of bias was assessed using the Cochrane risk-of-bias tool. Six RCTs met our inclusion criteria. Two RCTs found favorable effects of heat therapy on menstrual pain compared with unheated placebo therapy. Three RCTs found favorable effects of heating pads on menstrual pain compared with analgesic medication (n = 274; SMD −0.72; 95% confidence interval −0.97 to −0.48; P < 0.001; two studies). One RCT showed beneficial effects of heat therapy on menstrual pain compared with no treatment (n = 132; MD −4.04 VAS; 95% CI −4.88 to −3.20; P < 0.001). However, these results are based on relatively few trials with small sample sizes. Our review provided suggestive evidence of the effectiveness of heat therapy for primary dysmenorrhea, but rigorous high-quality trials are still needed to provide robust evidence.

Description of included trials: After removing duplicates, 1052 studies were screened and 15 full-text articles were assessed for eligibility. Three studies that used moxibustion were excluded because it delivers heat and excites the nervous system by acupoint stimulation. Two observational studies, one summary, and one trial protocol were also excluded. One study compared infrared heat to hot packs, and the other study was conducted in a non-randomized setting. Therefore, six RCTs were ultimately included in the analysis (Fig. 1). The characteristics of the included studies are summarized in Table 1. Two RCTs were conducted in America, and one RCT each was conducted in Iran, Korea, Taiwan, and Turkey. All of the studies were published in peer-reviewed journals. Four studies used a heating device, such as a patch or wrap, and two studies used a ceramic belt emitting far-infrared radiation (FIR). Details of the heat treatment are listed in Table 2. Akin et al. (2001 and 2004) reported only the mean value or the mean value and standard error of the reduction in pain scores. Furthermore, the exact number of participants in the intervention and control groups was unclear. Akin et al. (2004) reported that 357 participants finished the trial, and 11 participants were excluded; however, they finally analyzed 344 participants. Ke et al. just reported pain scores using figures with no numerical values. One of the authors (JJ) contacted the corresponding authors by electronic mail to request additional information, but the authors replied either that they had no raw data or did not respond. Therefore, meta-analyses were performed using the other two studies that compared a heat patch with an analgesic. Another study that compared the FIR belt with a heat pack with a placebo belt with a heat pack was reported separately. We used data from the first menstrual cycle after treatment, with the exception of one study, which reported a baseline difference in pain intensity during the first menstrual cycle.

Risk of bias in the included studies: The risk of bias in studies involving random sequence generation and blinding of outcome assessment was low in 33% of the trials (2/6) and unclear in 66% of the trials (4/6). The risk of bias for allocation concealment was low in 33% of the trials (2/6), unclear in 50% of the trials (3/6), and high in 17% of the trials (1/6). The risk of bias in blinding the participants and personnel was high in 33% of the studies (2/6) and low in 66% of the studies (4/6). There was a low risk of bias of incomplete outcome data, selective reporting, and other sources in all studies. Figure 2 summarizes the risks of bias and Appendix S2 provides the authors’ judgments on the risk of bias.

Outcome measures / Self-reported pain severity: Each of the six RCTs measured pain severity to identify the effects of heat therapy on alleviating menstrual pain. The meta-analysis of two studies showed that the heat patch had a more favorable effect on the severity of menstrual pain compared with analgesic medication (n = 274; SMD −0.72; 95% confidence interval [CI] −0.97 to −0.48; P < 0.001) and no treatment (n = 132; MD −4.04 VAS; 95% CI −4.88 to −3.20; P < 0.001) (Fig. 3). In two studies, Akin et al. reported that the heat patch demonstrated significant menstrual pain relief compared with unheated placebo therapy or acetaminophen. However, concurrent use of the heat patch and ibuprofen produced similar pain relief as the combined use of the unheated placebo patch and ibuprofen. Ke et al. showed a general trend towards a lower pain score in the FIR-belt group compared with the placebo-belt or blank group. Lee et al. found that the FIR belt with a heat pack and the placebo belt with a heat pack had similar effects on pain relief (n = 104; MD −0.73 maximal VAS; 95% CI −1.56 to 0.10; P = 0.08) (Fig. 3). However, we found a significantly greater effect on pain relief in the FIR-belt group (maximal VAS: 5.08 ± 2.24) compared with the placebo-belt group (maximal VAS 6.47 ± 2.24) in the two menstrual cycles immediately following the end of treatment (P = 0.002).

Validated pain questionnaires: One study used a shortened revision of the McGill questionnaire, which has proven validity and reliability, to measure pain. Pain severity was measured using 11 phrases describing sensory pain (0 to 33 points) and 4 phrases describing emotional pain (0 to 12 points). However, there were no significant differences between the heat patch and ibuprofen in terms of sensual and emotional pain (sensual pain MD = −0.04; 95% CI −2.25 to 2.17; P > 0.05; emotional pain MD = −0.50; 95% CI −1.40 to 0.40; P > 0.05).

Validated quality-of-life questionnaires: One study measured menstrual symptom severity (menstrual quality of life) using a standardized 17-item questionnaire. The four core symptom clusters (pain, negative affect, water retention, and food) were derived from a previous study. The heat-wrap group reported less severe menstrual symptoms (0.91 ± 0.49) compared with the acetaminophen group (0.99 ± 0.49); however, the difference was not statistically significant (MD = −0.08; 95% CI −0.19 to 0.03; P = 0.065). The pain cluster score (lower abdominal cramping, low backache, and generalized aches/pains) of the heat-wrap group (1.20 ± 0.74) was significantly lower than that of the acetaminophen group (1.35 ± 0.73; MD = −0.15; 95% CI −0.32 to 0.02; P = 0.040). Additionally, the heat-wrap group had significantly lower scores regarding mood swings (P = 0.046), fatigue (P = 0.012), and lower abdominal cramping (P = 0.008).

Adverse effects: Among the six studies, three noted if AEs were associated with treatment. Specifically, two studies reported AEs, and one study had no reported AEs. Mild conjunctivitis and moderate application site reactions occurred in the heat-wrap group, whereas moderate headache, rhinitis, and upper respiratory infection and severe anxiety occurred in the acetaminophen group. The frequency of AEs, including first-degree burns and itching, in the group that used the FIR belt with the heat pack was equal to that of the group that used the placebo belt with the heat pack. All AEs disappeared within a few days without treatment. There were no serious AEs, and no clinically relevant changes in vital signs; no patient discontinued the clinical trial due to an AE.