Comparative effectiveness of exercise interventions for primary dysmenorrhea: a systematic review and network meta-analysis

Abstract

Background

Exercise is increasingly being promoted as an effective treatment for primary dysmenorrhea (PD). This study aims to conduct a comprehensive network meta-analysis (NMA) of randomized controlled trials to identify the optimal types and dosages of exercise for managing PD in women.

Methods

Adhering to PRISMA-NMA guidelines, we systematically reviewed RCTs from the Cochrane Library, Web of Science, PubMed, and Embase databases up to May 23, 2024. Data analysis was performed using ‘GEMTC’ and ‘BUGSnet’ packages within a Bayesian framework in R and a hierarchy of exercise treatments was also calculated using surface under the cumulative ranking curve (SUCRA) values. Subgroup analyses were conducted to identify the most effective exercise regimens, including duration, frequency, and volume of the exercise interventions.

Results

Forty-nine studies representing 3,129 participants (1,640 exercises and 1,489 controls) were included. The results showed that all exercise interventions significantly reduced menstrual pain of the PD patients. Of six exercise intervention modalities based on the study ranked effectiveness, statistically significant reductions in pain intensity were found for resistance exercise and multi-component exercise. Multi-component exercise and stretching exercise were ranked best for menstrual symptoms, while core-strengthening exercise and multi-component exercise had the greatest impact on reducing pain duration. Significant and clinically important reductions or reliefs in pain occurred with 4 to 8 weeks of exercise training from all exercises, with resistance exercise showing the best efficacy when the duration exceeded 8 weeks, followed by multi-component exercise and mind-body exercise. Multi-component exercise and aerobic exercise with 1 to 3 sessions per week induced greater benefit in performance improvements, while resistance exercise with increased frequency showed the enhanced performance. Resistance exercise could elicit better efficacy within` 30-minute training duration, and multi-component exercise was ranked the best if such a training over 30 min.

Conclusion

This study provided quantitative insight into efficacy and effectiveness of exercise interventions on PD treatments. All six different exercises are associated with positive influence on PD management. Our study indicates that this exercise training induced adaptation may have therapeutic benefits for PD patients; however, such alterations and improvements are affected by exercise regiments.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12905-024-03453-w.

Background

Primary dysmenorrhea (PD), a prevalent gynecological condition that affects an estimated 16–91% of women of reproductive age [1]. It is characterized by painful uterine cramps during menstruation in the absence of any discernible pelvic pathology [2, 3]. The symptoms associated with PD, including menstrual pain, low back pain, fatigue, and muscle stiffness, often lead to absenteeism from school and work [2]. The global impact of PD causes a loss of approximately 600 million working hours and 2 billion US dollars annually [4]. Consequently, addressing the effects of PD on women’s life quality and their productivity is crucial.

Since the precise end-organ pathology of PD remains elusive, its pathophysiological mechanism is still not fully understood. Consequently, clinical treatment for PD often mirrors that of endometriosis, adenomyosis, and other disorders that manifest with menstrual pain, adopting a conservative approach. Recommended therapies to alleviate PD pain include administration of non-steroidal anti-inflammatory drugs (NSAIDs), contraceptives, analgesics, as well as massage and exercise. However, given the potential side effects associated with synthetic drugs, such as nausea, digestive issues, diarrhea, and renal complications [5, 6], exercise has emerged as a viable complementary and alternative therapy for menstrual pain management over the past three decades [7, 8]. Lorzadeh et al. reported that a 12-week regimen of core strength exercises effectively mitigated menstrual pain symptoms [9], including discomfort, anxiety, and sleep disturbances, among non-athlete female students. Similarly, a 12-week, 3 times per week regimen of aquatic exercise has been found to significantly reduce the severity and duration of PD pain in female college students [10]. Additionally, mind-body exercise like yoga and Pilates have gained popularity for their perceived benefits in managing menstrual symptoms [11, 12]. However, Shafaie et al. reported no significant difference between athletes and non-athlete female college students with regard to menstrual pain, premenstrual syndrome prevalence and symptoms [13], indicating the scholarly discourse remains divided on the efficacy of exercise for alleviating PD pain and relevant symptoms.

The lack of standardized measuring tools and detailed information about specific exercise programs for alleviating menstrual pain has contributed to these discrepancies. Traditional meta-analyses have struggled to reconcile these differences due to high heterogeneity among exercise programs. For example, there was an absence of an established hierarchy for determining which types of exercise (aerobic, strength training or both, mind-body exercise or mixed) might be best for beneficial to outcomes of PD changes based on both direct and indirect evidence. Network meta-analysis (NMA) is an innovative mathematical approach that facilitates the synthesis of indirect and direct comparisons to evaluate the effects of different exercise interventions while preserving randomization in individual trials [14–16]. Additionally, NMA allows for ranking interventions based on specific outcomes and demonstrates the probability of efficacy, which is essential for informing clinical guidelines and decision-making [17].

To the best of the authors’ knowledge, no previous network meta-analysis has examined the effects of aerobic, strength, resistance, mind-body training, or the combined aerobic and strength training on comparative effectiveness of exercise interventions for primary dysmenorrhea. Therefore, the primary objective of this study was to conduct a comprehensive systematic review with network meta-analysis of randomized trials to: (1) determine the effects of different exercise interventions (aerobic, core-strengthening, resistance, multi-component exercise, etc.) on PD treatments, and (2) identify a hierarchy of exercise interventions (aerobic, core-strengthening, resistance, multi-component exercise, etc.) in reducing PD pain intensity, duration, and associated menstrual symptoms for women with PD, and provide a grounded basis on recommendations for clinical usage and practice.

Methods

This study was performed and reported according to the guidelines provided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) protocol [18]. Given the systematic review nature of this research, ethics review board approval and participant informed consent was not required.

Data sources and search strategy

We conducted a comprehensive search of relevant articles published in English across the Cochrane Library, Embase, PubMed, and Web of Science up to May 23, 2024. The search strategy was developed by ZQY and ZY, incorporating advice from content expert and existing literature. HGY developed the search strategies using topic-related keywords, managing the search results across database, while CWJ refined the strategies. Details of the search terms and strategies are available in Supplementary Part A (Table S1-4). ZQY also manually reviewed reference lists to identify any potential additional studies eligible for inclusion in our analysis.

Inclusion criteria and outcome

The eligibility of original studies for inclusion was determined based on the following criteria: (1) RCTs conducted with women experiencing PD; (2) RCTs that compare two or more exercise interventions that fit predetermined categories, or that compare exercise interventions to non-exercise interventions or active medical treatments; (3) RCTs with an intervention period of at least one menstrual cycle or longer; (4) RCTs that assess outcomes related to menstrual pain intensity using validated measurement tool (Table S5).

Exercise interventions were categorized into 6 types [19, 20]: mind-body exercise (MBE), stretching exercise (SE), aerobic exercise (AE), core-strengthening exercise (CE), resistance exercise (RE), and multi-component exercise (ME), based on their clinical relevance to PD. We included 2 comparison groups: non-exercise intervention (NE) and active control (AC), detailed in Supplementary Part B (Table S5). Primary outcomes focused on pain intensity assessed using the Visual Analogue Scale (VAS), the Numerical Pain Rating Scale (NPRS), and the McGill Pain Questionnaire (MPQ). Secondary outcomes, which included focused on pain duration and menstrual symptoms, were evaluated using the Menstrual Distress Questionnaire (MDQ), Menstrual Symptoms Questionnaire (MSQ), Premenstrual Syndrome Scale (PMSS), and Premenstrual Symptoms Screening Tools (PSST) [21–23].

Study selection and data extraction

Duplicate studies were removed. ZQY and ZY independently screened titles and abstracts with full texts reviewed where eligibility was unclear. Studies were included if they met the predetermined criteria (Table S5). Discrepancies were resolved through discussion or expert consultation.

Two researchers (ZQY and ZY) conducted data extraction followed Cochrane guidelines [24], focusing on study and participant characteristics, exercise parameters (type, duration, volume, intensity, and frequency), comparator information, and outcome results.

Risk-of-bias assessment

Risk of bias was assessed using the Cochrane Risk of Bias Tool 2.0 for randomized trials (2019) by two pairs of authors (ZQY & ZY and HGY & CWJ) [25]. This tool evaluates five critical areas: (1) Randomization process; (2) Deviations from the intended interventions; (3) Missing outcome data; (4) Measurement of outcome; (5) Selection of the reporting results. We contacted the corresponding author(s) for missing or insufficient information, and if accurate data could not be obtained, the study was excluded. Each study was rated as low, some concern, or high risk of bias with disputes resolved through discussions.

Data synthesis

1. Network meta-analysis

Bayesian contrast-based multilevel NMA models [26] and Markov Chain Monte Carlo (MCMC) simulation were used to analyze treatment comparisons, calculating stable distribution probabilities and the area under the posterior distribution curve [26, 27]. The NMA was conducted using the ‘GEMTC’ and ‘BUGSnet’ packages in R [27–30].

First, we applied the random-effects model to address the clinical and methodological diversity within studies to provide more conservative confidence intervals for pooled point estimates. Second, we determined the appropriate model (consistency or inconsistency) by comparing the values of deviance information criteria (DIC) [30, 31]. A smaller DIC indicates a better model fit, and a difference of less than three points between models suggests no significant discrepancies [31]. If there is no global inconsistency, we performed node-splitting analyses to ensure local consistency in comparisons [32]. To assess treatment effectiveness, we generated ranking results from MCMC simulation, displayed in surface under the cumulative ranking (SUCRA) plots [33], where higher SUCRA value suggests better efficacy. We also presented the results of all pairwise comparisons in ranking order in league table heat plots to demonstrate the effectiveness of treatments visually [30, 33].

• (2)Convergence and consistency

The convergence of the MCMC was evaluated using the Gelman-Rubin diagnostic, which calculated the potential scale reduction factor (PSRF) to determine if the mean estimates had converged [27]. A PSRF value close to 1 suggests that approximate convergence has been achieved, although a more conservative threshold of 1.05 is often recommended in practice [28].

In addition to addressing the inherent heterogeneity in comparisons, we also addressed inconsistency that might arise in traditional meta-analysis but are detectable only through the NMA. We employed the node-splitting method to examine the contribution of each dataset to deviations and to evaluate the fit of the NMA model [32]. We calculated p-values for each comparison to identify inconsistencies between direct and indirect comparisons. If the p-value from the node-splitting analysis is less than 0.05, it suggests significant inconsistency within a specific comparison [30].

• (3)Subgroup analysis

In order to create a detailed exercise recommendation for alleviating menstrual pain, we conducted subgroup analyses focusing on crucial aspects of exercise interventions: intensity, duration, frequency, and volume [34, 35]. Due to insufficient data on exercise intensity in most studies, we limited out subgroup analyses to the duration, frequency, and volume. We categorized studies into groups based on duration (4 to 8 weeks or longer than 8 weeks), frequency (1 to 3 times per week or more than 3 times a week), and volume (30 min per session or longer). Following this, we performed separate network meta-analyses within these subgroups to determine the optimal exercise parameters for effective pain relief.

Results

Study selection and characteristics

A total of 1,471 articles were retrieved from four databases. After removing duplicate articles, 1,435 remained. Following the review of titles and abstracts, 106 articles were retained. Upon thorough examination of the full texts, 49 eligible RCTs were ultimately included in this NMA analysis, consisting of 46 two-armed RCTs and 3 three-armed RCTs (Fig. 1). Detailed explanations for exclusions and references can be found in the Supplementary Part D (Table S7).

Fig. 1.

PRISMA diagram of study search and selection

In the included studies, Asia, particularly South Asia, had the highest number of publications, with 19 studies from India and Pakistan. The total number of participants was 3,129, with a mean age ranging from 14 to 34 years. There were 1,640 participants in the exercise groups and 1,489 in the control groups. Sample sizes in the exercise groups ranged from 11 to 97, while control groups varied from 8 to 90. No significant differences in demographic factors were observed at the baseline (P > 0.05). Pain intensity was primarily measured using the VAS in 41 studies, the NPRS in 6 studies, and the MPQ in 2 studies. Additionally, 11 studies reported on menstrual pain duration, and 16 studies assessed the impact of exercise on various menstrual symptoms beyond the primary outcomes. Detailed characteristics of the included studies are presented in Table 1.

Table 1.

Characteristics of included studies

Author (year)	Country	Study Design	Participants	N	Mean age	Comparison	EG (time, frequency, duration)	CG/EG (time, frequency, duration)	Outcomes
Abbas et al., 2023 [36]	Egypt	RCT	Post-acute COVID-19 women (aged from 18 to 25 years) suffering from primary dysmenorrhea	30	24.5 ± 2.60	ME/NE	EG (n = 15): 10 reps/exercise, 30 min/session, 2 sessions/day, 3 days/week, 4 weeks	CG (n = 15): Non-exercise intervention	VAS
Abdelaziz et al., 2020 [37]	Egypt	RCT	Female participants (aged from 14–20 years) had main complaints that were pain and cramping during menstruation	60	17.645 ± 0.93	MBE/AC	EG (n = 30): 30 min/session, 3 sessions/week, 12 weeks	CG (n = 30): Active control (Kinesiotaping)	VAS
Agrawal and Ahmed, 2021 [38]	Maharashtra	RCT	Female students between the age group 16–25 years who were unmarried with regular menstrual cycle, and had moderate to severe dysmenorrhea.	60	NA	SE/CE	EG-1 (n = 30): 5 reps + holding/exercise, 1 session/day, 6 sessions/week, 8 weeks.	EG-2 (n = 30): 10 reps + holding/exercise, 1 session/day, 6 days/week, 8 weeks.	NPRS
Akbaş and Erdem, 2019 [39]	Turkey	RCT	Female participants (aged 18–25 years) with primary dysmenorrhea (at least 4 on a 10 cm VAS)	37	21.15 ± 1.51	AE/NE	EG (n = 18): 50 min/session, 3 sessions/week, 4 weeks	CG (n = 19): Non-exercise intervention	VAS
Aksu and Vefikuluçay, 2024 [40]	Turkey	RCT	Female university students (18 years, with regular menstrual periods) were diagnosed with primary dysmenorrhea	60	20.05 ± 1.20	MBE/NE	EG (n = 30): 60 min/session, 2 sessions/week, 12 weeks	CG (n = 30): Non-exercise intervention	VAS
Amreen et al., 2013 [41]	India	Three-armed RCT	Female students (aged between 19–25 years) were screened for primary dysmenorrhea	24	21.08 ± 1.282	RE/AC	EG-1 & EG-2 (n = 16): 30 contraction/set, 3 sets/session, 3–4 sessions/week, 8 weeks	CG (n = 8): Active control (10 min hot pack)	VAS
Arora et al., 2014 [42]	Navi Mumbai	RCT	Young female participants (aged from 19–24 years) with primary dysmenorrhea	60	20.565 ± 1.516	AE/NE	EG (n = 30): 50 min/session, 3–5 sessions/week, 12 weeks	CG(n = 30): Non-exercise intervention	VAS
Azima et al., 2015 [43]	Iran	RCT	Students residing in dormitories of Shiraz University, who were majoring in nonmedical fields and had primary dysmenorrhea	68	20.91 ± 1.15	CE/NE	EG (n = 34): 10 times/session, 2 sessions/day, 5 days/week, 8 weeks	CG (n = 34): Non-exercise intervention	VAS
Behbahani et al., 2016 [44]	Iran	RCT	Single female below 25 years old, having primary dysmenorrhea with pain intensity of four and above based on VAS	80	20.25 ± 1.52	CE/AC	EG (n = 40): 8 weeks	CG (n = 40): Active control (Routine treatment for dysmenorrhea)	MPQ
Berde, S.D. et al., 2019 [45]	India	RCT	Female college students with primary dysmenorrhea	50	NA	CE/AE	EG-1 (n = 25): 4 days/week, 8 weeks	EG-2 (n = 25): 4 days/week, 8 weeks	VAS
Boztas Elverisli et al., 2022 [46]	Turkey	Five-armed RCT	Female participants (aged from 18 to 30 years) with positive primary dysmenorrhea diagnosis	46	20.335 ± 1.64	ME/NE	EG (n = 23): 60 min/session, 2 sessions/week, 12 weeks	CG (n = 23): Non-exercise intervention	VAS
Celik and Apay, 2021 [47]	Turkey	RCT	Female students who experienced dysmenorrhea according to the VAS (scored 5 points or more)	124	20.17 ± 1.41	SE/NE	EG (n = 64): 30 min/session, 3–7 times/week, 8 weeks	CG (n = 60): Non-exercise intervention	VAS
Chaudhuri et al., 2013 [48]	India	RCT	School girls with recurrent, spasmodic menstrual cramps (primary dysmenorrhea)	128 (112)	14	ME/AC	EG (n = 53): 10–15 min/session, 2 sessions/day, 7 days/week, 12 weeks	CG (n = 75): Active control (Hot water bottle)	VAS
Elbandrawy and Elhakk, 2021 [49]	Egypt	Three-armed RCT	Female participants (aged from 18–25 years) with primary dysmenorrhea	105	22.39 ± 1.89	AE/CE/NE	EG-1 (n = 35): 45 min/session, 3 sessions/week, 8 weeks EG-2 (n = 35): 5 reps + holding/exercise, 45 min/session, 3 sessions/week, 8 weeks	CG (n = 35): Non-exercise intervention	VAS
Fallah and Mirfeizi, 2017 [50]	Iran	Four-armed RCT	College students aged 15–18 years with primary dysmenorrhea	42	15.61 ± 0.94	SE/NE	EG (n = 22): 20 min/session, 3 sessions/week, 2 times/day, 8 weeks	CG (n = 20): Non-exercise intervention	VAS
Huang et al., 2022 [51]	China (Taiwan)	RCT	Females (aged from 18–40 years old), being afflicted with primary dysmenorrhea	30	21.05 ± 2.31	AE/NE	EG (n = 15): 30–35 min/session, 2 sessions/week, 10 weeks	CG (n = 15): Non-exercise intervention	VAS
Ibrahim et al., 2023 [52]	Saudi Arabia	RCT	Females (18–23 years) with primary dysmenorrhea	33	20.93 ± 1.25	SE/NE	EG (n = 22): 10 reps + holding/exercise, 30–45 min/session, 3 sessions/week, 4 weeks	CG (n = 11): Non-exercise intervention	VAS
Jaibunnisha et al., 2017 [53]	India	RCT	Female participants with regular menstrual cycle and diagnosed with primary dysmenorrhea	67	NA	SE/NE	EG (n = 33): 10 min/session, 1 session/day, 6 days/week, 8 weeks	CG (n = 34): Non-exercise intervention	NPRS
Kannan et al., 2019 [54]	New Zealand	RCT	Women (18–43 years) with primary dysmenorrhea	55	NA	AE/NE	EG (n = 35): 3 times/week, 28 weeks	CG (n = 35): Non-exercise intervention	VAS
Kaur et al., 2014 [55]	India	Three-armed RCT	Girls of age 19–25 years with primary dysmenorrhea were selected from lovely professional university and government college	105	NA	SE/MBE/NE	EG-1 & EG-2 (n = 70): 4 days/week twice for 10 min, 8 weeks	CG (n = 35): Non-exercise intervention	NPRS
Khare and Jain, 2015 [56]	India	RCT	Primary dysmenorrhea cases (aged 15–17 years)	30	NA	ME/NE	EG (n = 15): 30 min/session, 2 sessions/day, 3 times/week, 3 weeks	CG (n = 15): Non-exercise intervention	VAS
Kirca and Celik, 2023 [57]	Turkey	RCT	Female university students in the 3rd and 4th year (aged from 18 to 24 years) with primary dysmenorrhea, and have a VAS value of 6 and over	60	20.38 ± 0.48	MBE/NE	EG (n = 30): 60 min/session, one session/week, 12 weeks	CG (n = 30): Non-exercise intervention	VAS
Kirmizigil and Demiralp, 2020 [58]	Turkey	RCT	Women diagnosed with primary dysmenorrhea between the ages of 18 and 35	28	23 ± 1.92	ME/NE	EG (n = 14): 50 min/session, 3 times/week, 8 weeks.	CG (n = 14): Non-exercise intervention	VAS
M.I. Ortiz et al., 2015 [59]	Mexico	RCT	Sedentary female patients with primary dysmenorrhea aged from 18–22 years (pain intensity from 4 to 10 cm based on VAS)	160	20.25 ± 1.21	ME/NE	EG (n = 83): 50 min/session, 3 sessions/week, 12 weeks	CG (n = 77): Non-exercise intervention	VAS
Mintu Merin et al., 2020 [60]	India	RCT	Female participants (aged from 18–25 years) with any primary dysmenorrhea symptoms	50	21.86 ± 0.904	AE/CE	EG-1: 4 days/week, 8 weeks	EG-2: 10s holding/exercise, 12 times/set, 3 sets/day, 3 days/week, 8 weeks	VAS
Motahari-Tabari et al., 2017 [61]	Iran	RCT	Students living in the university dormitory who had moderate to severe primary dysmenorrhea for more than 50% of menstrual cycles lasting for at least one day and affected their daily activities	122	21.45 ± 2.00	SE/AC	EG (n = 61): 15 min/session, 3 sessions/week, 8 weeks	CG (n = 61): Active control (Mefenamic acid treatment)	VAS
Ozturk et al., 2023 [62]	Turkey	Three-armed RCT	University female students who scored menstrual pain 6 or higher on the VAS	43	19.84 ± 1.51	SE/NE	EG (n = 22): 3 times/day on the first 3 days of the menstrual cycle, 8 weeks (two menstrual cycles).	CG (n = 19): Non-exercise intervention	VAS
Pastor, S. et al., 2023 [63]	India	RCT	Female college students (aged 18–25 years) with primary dysmenorrhea	30	19 ± 1.5	CE/AE	EG-1 (n = 15): 5–12 repetitions/position, 40 min/session, 1 session/day, 4 days/week, 8 weeks	EG-2 (n = 15): 5–12 repetitions/position, 40 min/session, 1 session/day, 4 days/week, 8 weeks	NPRS
Patel et al., 2015 [64]	India	RCT	Female participants (aged from 17 to 25 years) with regular menstrual cycles and experienced moderate to severe primary dysmenorrhea.	120	21.32	SE/NE	EG (n = 60): 2 times/day, 3 days/week, 8 weeks	CG (n = 60): Non-exercise intervention	VAS
Qaisar and Abbas, 2023 [65]	Pakistan	RCT	Female participants (aged 20–30 years) with primary dysmenorrhea	24	27.32 ± 2.08	CE/AC	EG (n = 12): 20s holding×10 reps/position, 5-mins rest/set, 20 min/session, 3 sessions/week, 6 weeks	CG (n = 12): Active control (Abdominal strengthening exercise with TENS)	NPRS
Raja Laxmi V. et al., 2016 [66]	India	RCT	Female participants (aged from 18–25 years) with primary dysmenorrhea	60	NA	SE/CE	EG-1 (n = 30): 10 min/session, 2 sessions/day, 4 days/week, 7 weeks	EG-2 (n = 30): 20 min/session, 2 sessions/day, 4 days/week, 12 weeks	VAS
Rakhshaee, 2011 [67]	Iran	RCT	Female students with primary dysmenorrhea, 18–22 years old	92	20.67	MBE/NE	EG (n = 50): 20 min/session, at least 14 days of the menstrual cycle (luteal phase), 8 weeks	CG (n = 42): Non-exercise intervention	VAS
Rashid et al., 2019 [68]	Iran	RCT	Participants’ age ranges from 18 to 24 with mild to moderate dysmenorrhea during the last three periods according to the McGill pain scale (1 < score < = 6.6)	86	18.65 ± 0.64	AE/AC	EG (n = 43): 20–47 min/session, 3 sessions/week, 8 weeks	CG (n = 43): Active control (Two physical education classes once a week)	VAS
Rostami et al., 2006 [69]	Iran	RCT	Students had regular menstruation and severe primary dysmenorrhea	142	16.56 ± 1.12	ME/NE	EG (n = 97): 20 min/session, 2 sessions/day, 8 weeks	CG (n = 45): Non-exercise intervention	VAS
S Saleh et al., 2016 [70]	Egypt	Three-armed RCT	Participants experienced moderate to severe symptoms of dysmenorrhea	126	20.72 ± 1.15	SE/CE/NE	EG-1 (n = 44): 4 stretching exercises, 10 min/time, 3 times/day, 3 days/week, 8 weeks EG-2 (n = 44): 4 core strengthening exercises, 20 min/time, 3 times/day, 4 days/week, 8 weeks	CG (n = 38): Non-exercise intervention	VAS
Sakuma et al., 2012 [71]	Japan	RCT	Healthy females (aged 20–64 years)	98	33.61 ± 12.02	MBE/NE	EG (n = 67): one session/day, 7 days/week, 2 weeks	CG (n = 31): Non-exercise intervention	VAS
Samy et al., 2019 [72]	Egypt	RCT	Participants diagnosed with primary dysmenorrhea	98	21.47 ± 1.47	AE/NE	EG (n = 49): 60 min/session, 2 times/week, 8 weeks	CG (n = 49): Non-exercise intervention	VAS
Shah et al., 2016 [73]	India	RCT	Students from SPB Physiotherapy College	40	NA	SE/NE	EG (n = 20): 4 days/week, 8 weeks	CG (n = 20): Non-exercise intervention	VAS
Shahrjerdi et al., 2019 [74]	Iran	RCT	Non-athletic, unmarried girls, aged 18–25 years, who suffered from moderate to severe primary dysmenorrhea	34	22.07 ± 0.98	CE/NE	EG (n = 17): 10 reps/exercise, 45–60 min/session, 3 sessions/week, 8 weeks.	CG (n = 17): Non-exercise intervention	NPRS
Shirvani et al., 2017 [75]	Iran	RCT	Female students living in the dormitory with moderate to severe primary dysmenorrhea	122	21.46 ± 2.05	SE/AC	EG (n = 61): 15 min/session, 3 sessions/week, 8 weeks	CG (n = 61): Active control (250 mg ginger capsules)	VAS
Song and Kim, 2023 [76]	Republic of Korea	RCT	Young female participants (aged from 19–39 years) with primary dysmenorrhea	30	32.6 ± 4.18	MBE/NE	EG (n = 15): 50 min/session, 2 sessions/week, 12 weeks	CG (n = 15): Non-exercise intervention	VAS
Sudhakar, S. et al., 2017 [77]	India	RCT	Female participants with primary dysmenorrhea	30	20.45 ± 3.42	MBE/CE	EG-1 (n = 15): 60s holding×5 reps×20s rest/exercise, 1 session/day, 3 days/week, 12 weeks	EG-2 (n = 15): 10s holding/exercise, 12 reps/set, 3 sets/day, 3 days/week, 12 weeks	VAS
Susan et al., 2018 [78]	India	RCT	Adolescent girl (18–23 years) with primary dysmenorrhea	30	NA	ME/SE	EG-1 (n = 15): 45 min/session, 3 sessions/week, 4 weeks	EG-2 (n = 15): 10 min/session, 3 sessions/week, 4 weeks	VAS
Temizkan and Budak, 2021 [79]	Turkey	Three-armed RCT	Women (aged from 15–30 years) with primary dysmenorrhea	30	22.6 ± 1.99	AE/NE	EG (n = 15): 45 min/session, 3 sessions/week, 3 weeks	CG (n = 15): Non-exercise intervention	MPQ
Tharani et al., 2018 [80]	India	RCT	Girls aged 17–23 years with regular menstrual cycle and VAS scoring > 6 and DASS-21 scoring > 19	30	NA	SE/AE	EG-1 (n = 15): 45 min/day, 3 days/week (alternate days), 8 weeks	EG-2 (n = 15): 45 min/day, 3 days/week (alternate days), 8 weeks	VAS
Yang and Kim, 2016 [81]	Republic of Korea	RCT	Undergraduate nursing students with primary dysmenorrhea	36	21.06 ± 0.53	MBE/NE	EG (n = 18): 60 min/session, 1 session/week, 12 weeks.	CG (n = 18): Non-exercise intervention	VAS
Yonglitthipagon et al., 2017 [11]	Thailand	RCT	Non-athlete women with primary dysmenorrhea aged 18–22 years	34	19.89 ± 1.20	MBE/NE	EG (n = 17): 30 min/session, two sessions/week, 12 weeks	CG (n = 17): Non-exercise intervention	VAS
Yosri, M.M. et al., 2022 [82]	Egypt	RCT	Female participants (aged from 19–25 years) having normal menstrual cycles, who were diagnosed with primary dysmenorrhea.	120	19.85 ± 1.16	MBE/RE	EG-1 (n = 30): 4 yogic positions, 6 days/week, 8 weeks	EG-2 & EG-3 & EG-4 (n = 90): 4 yogic positions, 30 min of squatting exercise, 6 days/week, 8 weeks	VAS
ZAID et al., 2022 [83]	Malaysia	RCT	Female participants (aged from 18–29 years) with regular menstrual cycle and were diagnosed with primary dysmenorrhea	24	22.58 ± 0.83	CE/NE	EG (n = 12): 10 reps/session, 10 min/session, 2 sessions/day, 5 days/weeks, 8weeks	CG (n = 12): Non-exercise intervention	VAS

MBE Mind-body exercise, SE Stretching exercise, RE Resistance exercise, CE Core-strengthening exercise, AE Aerobic exercise, ME Multi-component exercise, NE Non-exercise intervention, AC Active control, EG Exercise group, CG Control group, VAS Visual analogue scale, NPRS Numerical pain rating scale, MPQ McGill pain questionnaire, RCT Randomized controlled trial, N Sample number, NA Not Available, rep Repetition

Risk of bias

All studies considered the risk of low bias in generating random sequences. Among the 49 articles, 34 were deemed to have a low risk of allocation concealment bias, while 15 did not specify their allocation concealment methods and were considered to have an uncertain risk. Only 1 study was assessed as having a high risk of bias due to the randomization process, potentially affecting the reliability of its results. Most studies demonstrated a low risk of bias regarding deviations from the intended intervention (46), missing outcome data (48), measurement of outcomes (48), and selection of the reported result (41). Overall, 27 studies were considered to have a low risk of bias, 21 were noted to have some concerns, and 1 was deemed to have a high risk of bias. In summary, based on the Cochrane Handbook, most studies face some challenges related to low-to-moderate risk of bias. Detailed risk of bias ratings for each study is available in Supplementary Part E (Figure S2-3).

NMA

The comprehensive NMA graphs representing interventions for PD are presented in Fig. 2. In these graphs, each node represents a specific type of intervention. The size of each node indicates the number of studies included for that intervention, while the thickness of the connecting lines signifies the number of direct comparisons between two interventions.

1. Pain intensity

Fig. 2.

A NMA graph for pain intensity. B NMA graph for menstrual symptoms. C NMA graph for duration of pain. Note: MBE: mind-body exercise; SE: stretching exercise; RE: resistance exercise; CE: core-strengthening exercise; AE: aerobic exercise; ME: multi-component exercise; NE: non-exercise intervention; AC: active control

The network graph illustrates all available comparisons of pain intensity from the included trials (Fig. 2A). The Bayesian NMA model analyzed eight interventions: 6 exercise types (MBE, SE, AE, CE, RE, and ME) and 2 control groups (NE and AC). All 6 types of exercise showed positive effects in reducing menstrual pain compared to NE, with mean differences (MDs) ranging from − 5.2 (95% Credible interval (95% CI) = -8.1 to -2.3) for RE to -2.4 (95% CI = -3.7 to -1.2) for AE. However, there was no significant difference between AC controls and exercises in pain relief. The comparative efficacy of different exercises in alleviating menstrual pain is presented via SUCRA plot and league heat table (Fig. 3). The SUCRA analysis indicated that RE had the highest probability of being the most effective exercise type (93.07%), followed by ME (77.28%), MBE (68.15%), AC (54.41%), SE (42.88%), CE (32.13%), AE (32.08%), and NE (0.01%).

Fig. 3.

League heat table (left) and SUCRA values of the treatment rankings (right) of the pain intensity network. The symbol ** in the figure indicates significant differences between the treatments (P < 0.05). The SUCRA values represent the rankings of the treatments, while the upper curve—each color corresponding to a different treatment—displays the highest efficacy relative to the other curves (treatments) below

• (2)Menstrual symptoms

The network graph depicting menstrual pain displayed comparisons from the included trials (Fig. 2B). 16 studies examined the impact of various interventions on menstrual symptoms, 6 exercise types (MBE, SE, AE, CE, RE, and ME), and 2 control groups (NE and AC). SE showed a statistically significant improvement in menstrual symptoms (MD = -19.6, 95%CI = -39.7 to -0.1). Notably, all exercises had larger effect sizes than AC (MD = -0.7, 95%CI = -28.1 to 27.6), which had a negligible effect on reducing menstrual symptoms. The comparative efficacy of different exercises in alleviating menstrual pain was presented via SUCRA plot and league heat table (Fig. 4). The SUCRA analysis revealed that RE had the highest probability of being the most effective exercise type (74.37%), followed by SE (74.23%), CE (70.60%), MBE (52.98%), AE (45.85%), ME (45.59%), AC (21.90%), and NE (14.48%).

Fig. 4.

League heat table (left) and SUCRA values of the treatment rankings (right) of the menstrual symptoms network. The symbol ** in the figure indicates that there were significant differences between the treatments (P < 0.05). The SUCRA values represent the rankings of the treatments, while the upper curve—each color corresponding to a different treatment—displays the highest efficacy relative to the other curves (treatments) below

• (3)Duration of pain

The network graph illustrating pain duration displayed all available comparisons from the included trials (Fig. 2C). In terms of pain duration, 11 studies reported outcomes involving 5 different exercises (MBE, SE, AE, CE, and ME) and 2 control groups (NE and AC). Overall, the results indicated that only several interventions effectively reduced the pain duration. Specifically, CE (MD = -5.7, 95%CI = -8.7 to -3.4), ME (MD = -5.2, 95%CI = -9.8 to -0.4), AE (MD = -4.8, 95%CI = -8.8 to -1.0), and SE (MD = -4.4, 95%CI = -8.1 to -0.8). Meanwhile, AC (MD = -4.0, 95%CI = -8.2 to 0.0) showed a moderate effect, and MBE demonstrated a negligible effect (MD = -0.3, 95%CI = -5.0 to 3.9). The comparative efficacy of different exercises in alleviating menstrual pain was presented through SUCRA plot and league heat table (Fig. 5). The SUCRA analysis revealed that CE had the highest probability of being the most effective exercise type (81.39%), followed by ME (71.16%), AE (67.17%), SE (57.47%), AC (50.52%), MBE (13.76%), and NE (8.53%).

Fig. 5.

League heat table (left) and SUCRA values of the treatment rankings (right) of the pain duration network. The symbol ** in the figure indicates that there were significant differences between the treatments (P < 0.05). The SUCRA values represent the rankings of the treatments, while the upper curve—each color corresponding to a different treatment—displays the highest efficacy relative to the other curves (treatments) below

Convergence and consistency

The Gelman-Rubi diagnostic results suggest that the MCMC algorithm has reached the posterior distribution and the Bayesian NMA model has converged, as indicated by the value of PSRF in all comparisons approaching 1.00 (Table S8-10).

Leverage plots comparing the goodness of fit between consistent and inconsistent models are shown in Fig. 6. These plots confirm that the consistent model provided a better fit, indicating the absence of global inconsistency. Node-splitting analysis for pain intensity revealed no local inconsistency within each loop (P > 0.05). However, slight local inconsistency was observed in secondary outcomes when comparing AE and CE for menstrual symptoms (P = 0.039), indicating some discrepancy between direct and indirect comparisons. Similarly, the node-splitting analysis for pain duration identified statistical inconsistency in the direct and indirect comparison between SE and NE (P = 0.016). The results of the node-splitting analysis are presented in the Supplementary Part F (Table S11-13).

Fig. 6.

A Consistency model fit for pain intensity. B Consistency model fit for menstrual symptoms. C Consistency model fit for duration of pain. The lower DIC indicates a better model fit, and DIC value < 3 suggests minimal significant differences between models. Note: pD: the effective number of parameters; _res: the posterior mean of the residual deviance; DIC: deviance information criterion

Subgroup analysis for pain intensity

1. Exercise duration

For exercises with a duration of 4 to 8 weeks, 6 types of exercise (MBE, SE, AE, CE, RE, and ME), and 2 control groups (NE and AC) were analyzed. All exercise interventions were statistically effective in reducing pain, with MDs ranging from − 5.6 (95%CI = -8.5 to -2.7) for RE to -2.8 (95%CI = -4.0 to -1.6) for CE. RE (93.29%) was found to be the most effective exercise for relieving menstrual pain, followed by ME (65.90%), MBE (63.70%), AC (55.85%), AE (55.58%), SE (35.68%), CE (30.01%), and NE (0.01%).

For durations exceeding 8 weeks, only 4 exercises (AE, CE, MBE, ME) and 2 control groups (NE and AC) were analyzed. Among these, ME (MD = -3.8, 95%CI = -7.0 to -0.5) and MBE (MD = -3.5, 95%CI = -5.6 to -1.3) showed significant reductions in pain intensity. AE showed minimal effect on menstrual pain (MD = -0.1, 95%CI = -3.0 to 2.7), while CE (MD = -3.6, 95%CI = -9.0 to 1.9) and AC (MD = -2.9, 95%CI = -7.0 to 1.1) had moderate but non-significant effects. SUCRA analysis suggested that ME (74.56%) and MBE (69.74%) were more beneficial for exercise interventions lasting more than 8 weeks, followed by CE (67.35%), AC (58.58%), and AE (17.04%). The comparative efficacy of these interventions on alleviating menstrual pain is detailed in Supplementary Part G (Figure S4-5).

• (2)Exercise frequency

For the subgroup of 1 to 3 sessions per week, involving 5 exercise types (MBE, SE, AE, CE, and ME) and 2 control groups (NE and AC), all exercises except CE (MD = -1.5, 95%CI = -3.4 to 0.5) showed significant pain relief compared to NE. The MDs ranged from − 4.2 (95%CI = -6.2 to -2.2) for ME to -2.8 (95%CI = -4.2 to -1.4) for SE. SUCRA analysis indicated that ME (87.78%) and AE (70.98%) had the highest probabilities of effectiveness when performed 1 to 3 times weekly, followed by MBE (62.68%), AC (55.10%), SE (51.56%), CE (20.73%), and NE (1.14%).

For exercises performed more than 3 times per week, involving 6 exercises (MBE, SE, AE, CE, RE, and ME) and 2 control groups (NE and AC), most exercises showed statistically positive effects on pain relief, with MDs ranging from − 5.3 (95%CI = -9.1 to -1.6) to -2.2 (95%CI = -4.3 to -0.2). AE presented a negligible effect (MD = -0.3, 95%CI = -2.8 to 2.3) compared to NE. SUCRA analysis illustrated that RE (91.45%), MBE (71.38%), and ME (65.70%) were the most effective, followed by AC (55.02%), SE (53.81%), CE (43.60%), AE (11.98%), and NE (7.07%). The comparative efficacy of these interventions on alleviating menstrual pain is presented in Supplementary Part G (Figure S6-7).

• (3)Exercise volume

For the exercise volume subgroups, which involved 6 exercise types (MBE, SE, AE, CE, RE, and ME) and 2 control groups (NE and AC). When each exercise session lasted less than 30 min, most exercises yielded statistically significant pain relief compared to NE, MDs ranging from − 6.3 (95%CI = -10.4 to -2.2) to -2.6 (95%CI = -4.2 to -1.0). AE showed a moderate but non-significant effect (MD = -2.1, 95%CI = -5.9 to 1.7). SUCRA analysis indicated that RE (94.75%) had the highest probability of reducing menstrual pain intensity for sessions lasting less than 30 min, followed by ME (63.12%), MBE (63.07%), AC (56.21%), SE (48.65%), CE (38.13%), AE (34.09%), and NE (1.95%).

When exercise sessions exceeded 30 min, ME (MD = -4.1, 95CI% = -7.2 to -1.0), MBE (MD = -3.5, 95%CI = -6.5 to -0.5) and AE (MD = -2.8, 95%CI = -5.1 to -0.7) were found to be statistically significant for pain relief. SUCRA analysis revealed that ME (72.03%) appeared to be the most effective type, followed by RE (68.62%), AC (65.33%), MBE (61.73%), AE (48.78%), SE (40.43%), CE (38.63%), and NE (4.44%). The comparative efficacy of different exercises on alleviating menstrual pain is presented in Supplementary Part G (Figure S8-9).

Discussion

Although numerous studies have explored the efficacy of various exercise interventions for PD, no tailored exercise regimen specifically for women with PD has been thoroughly investigated. This study provides a comprehensive comparison of different exercise types for reducing the intensity and duration of dysmenorrhea and alleviating menstrual symptoms. As the most recent and extensive network meta-analysis on this topic, our study offers valuable insight into the effectiveness of diverse exercise interventions for PD.

Summary of Key findings

To determine the optimal exercise regimen for PD, this NMA analyzed 8 interventions from 49 RCTs. All exercise interventions significantly reduced the menstrual pain intensity compared to non-exercise controls. RE and ME were particularly effective in alleviating menstrual pain, as measured by VAS. RE and SE were most effective for treating menstrual symptoms, while CE and ME were superior in reducing the duration of dysmenorrhea. Notably, our study found that exercise may offer benefits comparable to or even exceeding those of medical treatments for PD.

Regarding exercise duration, RE showed more benefits for pain relief when performed for 4 to 8 weeks, while ME and MBE remained effective beyond 8 weeks. For frequency, 1 to 3 sessions per week of ME or AE provided significant pain relief, with RE, MBE, and ME being more effective at higher frequencies. As for exercise volume, RE was more effective with sessions shorter than 30 min, while ME showed increased benefits when the volume exceeded 30 min, with AE also becoming more beneficial.

Comparison with previous literature

Previous traditional meta-analyses have also highlighted the benefits of exercise for PD, consistent with our findings [84–87]. However, these studies primarily included participants under 25 and were limited by methodological constraints and insufficient evidence, preventing comparisons of different types [35, 84, 85], or assessments of exercise versus traditional medical therapies [86, 87]. Pairwise meta-analyses also struggled with significant heterogeneity due to combining various exercise interventions [88]. Fortunately, the NMA approach utilized in this study may overcome these limitations.

To our knowledge, only one other NMA has examined the comparative efficacy of different exercises for PD, including 29 RCTs with 1,808 participants [89]. Similar to our study, it found that exercise could alleviate menstrual pain. However, it differed in several aspects, such as focusing on effects at 4 weeks and 8 weeks, and recommended SE as the optimal exercise for pain reduction. In contrast, our study identified RE and ME as more effective for menstrual pain relief across a broader range of exercise durations. Subgroup analysis further indicated that ME and MBE were among the top interventions for both short-term and long-term pain relief, whereas SE was less effective and even less no compared to medical treatments. These discrepancies may be attributed to the differences in exercise categories and the inclusion of various pain measurement tools beyond the VAS.

Exploring suitable Exercise modalities and dosages for PD

Adequate engagement in exercise can alleviate pain quality and intensity by modulating hormones and inflammatory cytokines [90, 91]. Several studies have suggested that different exercise dosages and types can produce varying analgesic effects through distinct mechanisms [92–94], highlighting the importance of exercise intensity, duration, and type in pain management [95]. RE entails substantial physical exertion and muscle strain [96], and either long-term or short-term RE training has shown its analgesic role in pain control [97, 98]. Long-term regular RE training may also contribute to neural plasticity associated with muscle adaptation and strength-gaining [99, 100]. Our study found that individuals with PD probably benefit more from RE when performed more frequently (> 3 sessions per week), with sessions lasting up to 30 min proving to be particularly effective.AE has also been recognized for its efficacy in pain management [101–103]. The effectiveness of AE is closely related to its dosage [103, 104], and maintaining moderate level of AE may yield better results than continuously increasing exercise doses [105]. Our finding suggested that engaging in at least 30 min of AE per session is beneficial for pain relief, aligning with previous research indicating that 40 min of AE can elevate brain-derived neurotrophic factor (BDNF) levels, which contributes to neuroprotection [106]. Moreover, our study revealed that MBE, ME, and CE demonstrated excellent applicability for long-term practice, implying that integrating these exercises into daily life could be practical. Therefore, appropriate participation in these exercises, either alone or in combination, not only helps in pain signals but also promotes muscle strengthening and overall health [97–100].

Although there is no definitive evidence on the most effective exercise regimen for PD, several qualitative analyses have offered some recommendations. Carroquino et al. suggested that an 8 to 12-week of exercise could help alleviate menstrual pain [85], while our study indicated that a 4- to 8-week duration of exercise may be sufficient to alleviate menstrual pain. Variations effects can be influenced by factors such as age, fitness level, and health condition [103, 107–109]. Younger individuals may experience more pronounced pain relief [109], and AE might have a greater impact on pain perception than RE [109]. In summary, characterizing different exercise types and dosages is essential for optimizing. Different exercises may provide specific advantages, and understanding their suitability can guide effective pain management strategies.

Strengths and limitations

The study’s strengths include its detailed classification of exercise modalities, use of multiple pain assessment tools (VAS, NPRS, MPQ), and application of Bayesian methods to handle complex models and small sample sizes, all of which enhance the reliability and robustness of the findings. Additionally, the detailed subgroup analyses provide practical insights into optimal exercise parameters for managing PD. However, the study has limitations, such as the small sample sizes of several included studies, the limited number of interventions assessed for secondary outcomes, and the focus on women aged 14 to 35, which may limit generalizability to other age groups. Moreover, the predominance of studies from Asian countries may affect the applicability of the results to other regions. While the study provides statistically significant findings, the lack of a well-defined minimal clinically important difference (MCID) for pain intensity in PD means that results should be interpreted with caution due to potential heterogeneity from clinical and methodological diversity [84, 110]. Future research should be conducted with a larger sample and include participants from diverse age groups and countries. Besides, future researchers should further determine the overall impacts of exercise on primary dysmenorrhea, including menstrual duration and associated symptoms. Collaborating with clinical practitioners will be essential to facilitate the implementation of exercise therapy in the treatment of primary dysmenorrhea. Moreover, high-quality RCTSs are required to further compare the exercise with specific parameters such as intensities, durations, and frequencies to explore the optimal exercise modalities for PD populations.

Conclusion

This NMA provides comprehensive evidence on the effectiveness of various exercise interventions for managing PD, highlighting that all examined exercise types significantly reduce menstrual pain intensity compared to NE controls. RE and ME were identified as the most effective for alleviating pain. Additionally, RE and SE have shown the best efficacy in improving menstrual symptoms, while CE and ME are more effective in reducing the duration of menstrual pain. The study also underscores that moderate-duration and frequency of exercise are crucial for optimal pain relief, with longer sessions and higher frequencies of RE, ME, and MBE offering additional advantages. The results suggest that tailored exercise regimens can be a viable alternative or complement to traditional medical treatments for PD, but further research is needed to refine exercise recommendations and enhance generalizability across different populations.

Abbreviations

PD

Primary dysmenorrhea

NSAIDs

Non-steroidal anti-inflammatory drugs

NMA

Network meta-analysis

RCT

Randomized controlled trial

PRISMA

Preferred reporting items for systematic reviews and meta-analyses

MBE

Mind-body exercise

SE

Stretching exercise

AE

Aerobic exercise

CE

Core-strengthening exercise

RE

Resistance exercise

ME

Multi-component exercise (ME)

NE

Non-exercise intervention

AC

Active control

VAS

Visual analogue scale

NPRS

Numerical pain rating scale

MPQ

McGill pain questionnaire

MDQ

Menstrual distress questionnaire

MSQ

Menstrual symptoms questionnaire

PSS

Premenstrual syndrome scale

PSST

Premenstrual symptoms screening tools

MCMC

Markov Chain Monte Carlo

DIC

Deviance information criteria

SUCRA

Surface under the cumulative ranking

PSRF

Potential scale reduction factor

MD

Mean difference

95%CI

95% Credible interval

BDNF

Brain-derived neurotrophic factor

MCID

Minimal clinically important difference

Authors’ contributions

ZQY and ZY contributed to the research and were responsible for conception and design. ZQY and CWJ were responsible for data analysis/interpretation and wrote the manuscript. HGY and ZY edited the manuscript and ZY supervised the entire process. All authors approved the final version.

Funding

This research was funded by the National Natural Science Foundation of China (NO.82102669).

Data availability

The original contributions presented in this study are included in the article and the Supplementary material. Further inquiries can be directed to the corresponding author.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.