A systematic review of exercise interventions on negative affect, pain, and fatigue in women with premenstrual syndrome

Abstract

Background

Premenstrual syndrome refers to a constellation of cyclical somatic, psychological, and behavioral changes that recur during the late luteal phase of the menstrual cycle. These recurring disturbances impair daily life and occupational performance, and their long-term accumulation may predispose individuals to more severe physical or psychological conditions.

Objective

We aimed to systematically evaluate the efficacy of exercise interventions for alleviating negative affect, pain, and fatigue in women diagnosed with premenstrual syndrome (PMS).

Design

A systematic review and meta-analysis was conducted.

Methods

Search of publications was conducted in Web of science, Embase, Pubmed, and The Cochrane Library. Treatment effects were defined as standardized mean difference and their 95% confidence intervals. Statistical heterogeneity was assessed using Crohan’s Q test and quantified using the I² index.

Results

Of the 20 articles selected, six applied running ; three, Pilate ; and eleven, yoga. The meta-analysis showed that exercise improved negative affect (P < 0.01), pain symptoms (P < 0.01), and fatigue symptoms (P < 0.01). Subgroup analyses revealed that running for 9–12 weeks at a frequency of three or more sessions per week appeared to be more effective in alleviating negative affect. Similarly, yoga practiced with the same duration and frequency (9–12 weeks and ≥ 3 sessions/week) appeared to be more effective in pain symptoms, while a program of 9–12 weeks appeared to be more effective for reducing fatigue.

Conclusion

Exercise has a ameliorative effect on negative affect, pain, and fatigue in women with PMS. Among the programs studied, running for 9–12 weeks at ≥ 3 sessions/week appeared to be more effective in negative affect, while yoga for 9–12 weeks at ≥ 3 sessions/week appeared to be more effective for pain relief, and yoga for 9–12 weeks appeared to be more effective for fatigue alleviation.

Prospero registration

CRD420251088581.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12905-025-04242-9.

Background

Premenstrual syndrome (PMS) refers to a constellation of cyclical somatic, psychological, and behavioral changes that recur during the late luteal phase of the menstrual cycle. It is highly prevalent among women of reproductive age and has substantial impacts on their physical and mental health [1]. According to the Global Burden of Disease study [2], the number of PMS cases worldwide increased from 652.5 million in 1990 to 956.0 million in 2019, representing a 46.5% rise. Most women of reproductive age experience at least one emotional or physical symptom in the premenstrual phase [3]. PMS is typically accompanied by negative affect—such as depression, anxiety, and irritability—as well as somatic symptoms including pain and fatigue [4–6]. These recurring disturbances impair daily life and occupational performance, and their long-term accumulation may predispose individuals to more severe physical or psychological conditions. Thus, identifying effective strategies to alleviate PMS-related symptoms has become an urgent priority in women’s health research.

Current PMS management strategies mainly include pharmacological therapy, psychological intervention, and lifestyle modification [7, 8]. Pharmacological treatments, such as fluoxetine and other selective serotonin reuptake inhibitors, can partially relieve symptoms but often cause adverse effects including nausea and dizziness [7–10]. Cognitive behavioral therapy (CBT) has also demonstrated clinical efficacy but is limited by cost, time, and the availability of adequately trained professionals [8, 11]. In contrast, exercise intervention is characterized by high feasibility and minimal side effects, making it particularly suitable for patients who decline or cannot tolerate medication [8]. Accumulating evidence suggests that exercise promotes the secretion of neurotransmitters such as endorphins and dopamine, which help regulate mood and relieve stress [12–14]. Regular physical activity also modulates the synthesis of progesterone and estrogen and enhances the endogenous release of anti-inflammatory mediators [15–17]. Beyond these physiological effects, exercise contributes to overall health, fosters social interaction, and reduces the risk of depressive symptoms—all of which may synergistically alleviate the manifestations of PMS [5, 18, 19].

Previous studies have shown that yoga can effectively relieve PMS-related symptoms [20–22]. However, most prior research has focused on a single exercise modality, leaving uncertain whether other forms of exercise yield comparable benefits. Moreover, the dose–response relationships among exercise parameters—including type, duration, frequency, and intervention period—remain insufficiently clarified. To address these gaps, the present study conducted a meta-analysis of randomized controlled trials (RCTs) investigating the effects of exercise interventions on negative affect, pain, and fatigue in women with PMS. We further examined the dose–response characteristics of key exercise parameters to identify optimal intervention patterns and provide evidence-based guidance for clinical rehabilitation and individualized exercise prescriptions for PMS management.

Methods

This systematic review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [23]. Based on the International Classification of Functioning, Disability and Health (ICF) framework, the effects of exercise interventions on negative affect, pain, and fatigue in women with premenstrual syndrome (PMS) were analyzed in relation to intervention cycle, modality, duration, and frequency. We further explored potential dose–response patterns across these parameters to identify optimal intervention characteristics. The PICOS framework is presented in Table 1.

Table 1.

PICOS framework for the meta-analysis of exercise interventions on symptom severity in patients with premenstrual syndrome

Population	Intervention	Comparison	Outcome	Study
Patients with PMS	Intervention setting Hospital Community health service center University gymnasium Intervention providers Pilates and yoga instructors Intervention duration Intervention time Intervention frequency	Exercise group vs. control group Comparison of exercise intervention duration, type, time, and frequency	Negative Affect Pain Fatigue	RCT

This review was prospectively registered in the International Prospective Register of Systematic Reviews ( PROSPERO, registration number CRD420251088581).

Inclusion and exclusion criteria

Inclusion Criteria

1. Study design: Randomized controlled trials (RCTs).

2. Participants: Women diagnosed with PMS according to either the Diagnostic and Statistical Manual of Mental Disorders (DSM, American Psychiatric Association) or the International Society for Premenstrual Disorders (ISPMD) criteria, irrespective of race or nationality. Participants were required to have regular menstrual cycles, no structured exercise within the previous three months, and no ongoing treatment with chemical agents, herbal preparations, or oral contraceptives.

3. Intervention: Exercise interventions (e.g., running, yoga, Pilates) in the experimental group, control groups included blank controls, usual care, or placebo conditions.

4. Outcomes: The primary outcome was negative affect (e.g., anxiety, depression, irritability), assessed with validated instruments such as the Premenstrual Syndrome Scale, Beck Depression Inventory, Depression Anxiety Stress Scale-21 (DASS-21), Moos Menstrual Distress Questionnaire, and the Hamilton Anxiety Rating Scale. Secondary outcomes were pain and fatigue, measured using the McGill Pain Questionnaire, Visual Analogue Scale (VAS), and Borg Scale.

Exclusion Criteria

1. Studies including healthy women or participants with primary dysmenorrhea.

2. Trials with insufficient or unextractable data.

3. Multi-component interventions (e.g., exercise combined with diet or nutritional supplementation).

4. Control groups not involving blank or routine treatment conditions.

5. Outcomes inconsistent with this review’s endpoints.

6. Systematic reviews, conference abstracts, or theses.

7. Duplicate publications or low-quality studies.

Search strategy

Two independent reviewers systematically searched PubMed, Embase, Web of Science, and The Cochrane Library from database inception to 20 June 2025 for RCTs investigating exercise interventions on negative affect, pain, and fatigue among PMS patients. Reference lists of eligible studies were screened manually, and study authors were contacted when necessary to obtain additional data.

	Search steps
PubMed	#1(exercise[MeSH Terms]) OR (walking[Title/Abstract] OR jogging[Title/Abstract] OR swimming[Title/Abstract] OR cycling[Title/Abstract] OR yoga[Title/Abstract] OR yogic[Title/Abstract] OR Pilates[Title/Abstract] OR resistance[Title/Abstract] OR Strength[Title/Abstract] OR dance[Title/Abstract] OR hiking[Title/Abstract] OR Tai Chi[Title/Abstract] OR exercise[Title/Abstract] ORaerobic[Title/Abstract] OR train[Title/Abstract] OR training[Title/Abstract] OR movement[Title/Abstract] OR Physical activity[Title/Abstract] OR running[Title/Abstract] OR Sports[Title/Abstract] OR Baduanjin[Title/Abstract] OR bike[Title/Abstract] OR workout[Title/Abstract] OR fitness[Title/Abstract] OR exertion[Title/Abstract] OR skipping[Title/Abstract] OR climbing[Title/Abstract] OR bicycling[Title/Abstract]) #2(premenstrual syndrome[MeSH Terms]) OR (Premenstrual[Title/Abstract] OR pre-menstrual[Title/Abstract] OR Premenstrual Dysphoric Disorder[Title/Abstract] OR Late Luteal Phase Dysphoric Disorder[Title/Abstract]) #3 (Randomized controlled trial[Publication Type]) OR(Randomized[Title/Abstract] OR controlled[Title/Abstract] OR trial[Title/Abstract]) #4 #1 AND #2 AND #3
The Cochrane Library	#1 (“walking” OR “jogging” OR “swimming” OR “cycling” OR “yoga” OR “yogic” OR “Pilates” OR “resistance” OR “Strength” OR “dance” OR “hiking” OR “Tai Chi” OR “exercise” OR “aerobic” OR “train” OR “training” OR “movement” OR “Physical activity” OR “running” OR “Sports” OR “Baduanjin” OR “bike” OR “workout” OR “fitness” OR “exertion” OR “skipping” OR “climbing” OR “bicycling” )Title Abstract Keyword AND “Exercise”[MeSH Terms] #2 (“Premenstrual Syndrome” OR “Premenstrual” OR “pre-menstrual” OR “Premenstrual Dysphoric Disorder” OR “Late Luteal Phase Dysphoric Disorder”)Title Abstract Keyword AND “Premenstrual Syndrome”[MeSH Terms] #3 (“Randomized controlled trial”)Publication Type OR ༈“Randomized”OR“controlled”OR“trial”༉Title Abstract Keyword #4 #1 AND #2 AND #3
Web of Science	#1 TS=(“walking” OR “jogging” OR “swimming” OR “cycling” OR “yoga” OR “yogic” OR “Pilates” OR “resistance” OR “Strength” OR “dance” OR “hiking” OR “Tai Chi” OR “exercise” OR “aerobic” OR “train” OR “training” OR “movement” OR “Physical activity” OR “running” OR “Sports” OR “Baduanjin” OR “bike” OR “workout” OR “fitness” OR “exertion” OR “skipping” OR “climbing” OR “bicycling” ) #2 TS=(“Premenstrual Syndrome” OR “Premenstrual” OR “pre-menstrual” OR “Premenstrual Dysphoric Disorder” OR “Late Luteal Phase Dysphoric Disorder”) #3 TS=(“Randomized controlled trial” OR “Randomized” OR “controlled” OR “trial”) #4 #1 AND #2 AND #3
Embase	#1“exercise”[exp]OR “walking”[ab, ti] OR “jogging”[ab, ti] OR “swimming”[ab, ti] OR “cycling”[ab, ti] OR “yoga”[ab, ti]OR “yogic”[ab, ti] OR “pilates”[ab, ti] OR “resistance”[ab, ti] OR “strength”[ab, ti] OR “dance”[ab, ti] OR “hiking”[ab, ti] OR “tai chi”[ab, ti] OR “exercise”[ab, ti] OR “aerobic”[ab, ti] OR “train”[ab, ti] OR “training”[ab, ti] OR “movement”[ab, ti] OR “physical activity”[ab, ti] OR “running”[ab, ti] OR “sport”[ab, ti] OR “baduanjin”[ab, ti] OR “bike”[ab, ti] OR “workout”[ab, ti] OR “fitness”[ab, ti] OR “exertion”[ab, ti] OR “skipping”[ab, ti] OR “climbing”[ab, ti] OR “bicycling”[ab, ti] #2“premenstrual syndrome”[exp]OR“premenstrual dysphoric disorder”[ab, ti]OR“pre- menstruali”[ab, ti]OR“premenstrual dysphoric disorder”[ab, ti]OR“late luteal phase dysphoric disorder”[ab, ti] #3“Randomized controlled trial”[exp]OR“Randomized”[ab, ti]OR“Controlled”[ab, ti]OR“Trial”[ab, ti] #4 #1 AND #2 AND #3

Study selection and data extraction

Study Selection

All retrieved records were imported into EndNote for duplicate removal. Two reviewers independently screened titles and abstracts according to the eligibility criteria, followed by full-text review of potentially relevant studies. Disagreements were resolved through discussion with a third reviewer until consensus was reached.

Data Extraction

Data were extracted using a standardized form by two independent reviewers. When data were incomplete or unclear, authors were contacted for clarification. Discrepancies were resolved by consultation with a third reviewer. Extracted information included:

1. Basic study information (author, publication year).

2. Study characteristics (sample size, baseline features, intervention details—exercise duration, frequency, period, rest intervals).

3. Outcome measures for negative affect (anxiety, depression, irritability), pain, and fatigue.

Quality assessment

Methodological quality was assessed using the Physiotherapy Evidence Database (PEDro) scale [24], which evaluates ten criteria: random allocation, allocation concealment, baseline comparability, blinding of participants, therapists, and assessors, attrition ≤ 15%, intention-to-treat (ITT) analysis, between-group comparison, and point estimates with variability data. One point was awarded for each criterion met (maximum = 10, eligibility not scored). Scores < 4 indicated low quality, 4–5 moderate quality, 6–8 good quality, and 9–10 high quality.

Additionally, outcome-specific evidence quality was rated using the GRADE (Grading of Recommendations Assessment, Development, and Evaluation) system [25, 26], categorizing evidence as high, moderate, low, or very low. Two reviewers independently rated all outcomes, disagreements were resolved by a third reviewer.

Statistical analysis

All analyses were performed using Stata 15.1 (StataCorp, College Station, TX, USA). Pooled effect sizes, heterogeneity, publication bias, and subgroup analyses were computed. Because continuous outcomes were reported using different scales, results were standardized using the standardized mean difference (SMD) with 95% confidence intervals (CI) [27]. A P < 0.05 indicated statistical significance.

SMD values of < 0.2, 0.2 to < 0.5, 0.5 to < 0.8, and ≥ 0.8 were interpreted as representing trivial, small, moderate, and large effect sizes, respectively, Heterogeneity was assessed using Cochran’s Q and the I² statistic. When P ≥ 0.1 and I² ≤ 50%, a fixed-effects model was applied, otherwise (P ≤ 0.1, I² ≥ 50%), a random-effects model was used. Heterogeneity was classified as very low (I² < 25%), low (25–50%), moderate (50–75%), or high (I² ≥ 75%) [28, 29].

For studies with multiple experimental or control groups, each distinct exercise or non-exercise comparison was treated as an independent dataset [30]. Publication bias was evaluated visually by funnel-plot symmetry and statistically using Egger’s test [31], if bias was detected, the non-parametric trim-and-fill method was used to assess result stability [32]. Sensitivity analysis was conducted by sequentially omitting each study to test the robustness of pooled outcomes [33].

Results

Literature search

A total of 3,789 potentially relevant articles were retrieved from four databases: PubMed (n = 100), Embase (n = 210), Web of Science (n = 2,537), and The Cochrane Library (n = 942). Five additional studies were identified through manual searching. After removing 485 duplicates, 63 records remained for title and abstract screening. Following full-text review, 20 RCTs met the inclusion criteria and were included in the meta-analysis [34–53].The study-selection process is illustrated in Fig. 1.

Fig. 1.

Flow diagram of study selection

Characteristics of included studies

Table 2 summarizes the characteristics of the included studies. The 20 trials involved 1,230 participants (603 in intervention groups, 627 in control groups), all diagnosed with PMS. Publication years ranged from 2012 to 2025. Exercise modalities in the intervention groups included running, Pilates, and yoga, while control groups received either no intervention or maintained normal daily activity.

Table 2.

Baseline characteristics of the included studies

Inclusion study	Nation/ Region	Number of cases(T/C)	Age(T/C)	Intervention type(T/C)	Intervention dose			Measurements
					intervention duration(min)	Frequency(T/W)	intervention period(W)
Hassan et al.2025 [50]	Iran	12/12	16.87 ± 0.39/17.2 ± 0.27	Running/ No arrangements	60	/	8	⑤
Zahra et al. 2018 [35]	Iran	35/30	24.06 ± 4.71/ 25.22 ± 4.41	Running/ Daily activities	30	3	8	⑥①②⑯
A.El-Lithy et al.2014 [48]	Egypt	15/15	17.80 ± 1.47/18.14 ± 1.51	Running + 50mgB6 + 1200mgCa/ 50mgB6 + 1200mgCa	25	3	12	③
Elvan et al. 2019 [51]	Turkey	35/32	19.28 ± 1.63/19.61 ± 2.11	Running/ No arrangements	30	3	12	⑪⑨
Mitra et al. 2012 [53]	Iran	T1:30 T2:30 C:30	T1: 27.1 ± 10.98 T2:26.8 ± 5.96 C:30.1 ± 12.15	Running/ No arrangements	50	3	12	④
Ayça et al. 2023 [41]	Turkey	30/29	20 ± 1.17/20.1 ± 2.35	Pilates/ No arrangements	60	2	8	⑪
Elif et al. 2020 [39]	Turkey	25/25	18.80 ± 0.86/ 19.16 ± 0.80	Pilates/ Daily activities	60	3	12	⑪
Şeyma et al. 2024 [45]	Turkey	23/23	25.69 ± 1.94/ 25.69 ± 1.84	Pilates/ No arrangements	45 ~ 55	2	8	⑦⑧⑪
Meryem et al. 2024 [44]	Turkey	39/41	20.56 ± 1.72	Yoga / No arrangements	42	2	8	⑨⑭⑧⑩
Mahin et al. 2017 [47]	Iran	26/28	30.179 ± 6.290/34.462 ± 5.368	Yoga / No arrangements	60	2	10	⑫
Didem et al. 2020 [37]	Turkey	50/54	19–25	Yoga / No arrangements	60	1	12	⑪⑩
Duygu et al. 2023 [42]	Turkey	30/30	31.17 ± 5.279/ 29.17 ± 4.632	Yoga / No arrangements	45	3	12	⑪⑩⑨
Ece Karali et al. 2024 [34]	Turkey	32/48	20.6 ± 0.7/ 20.5 ± 1.1	Yoga / No arrangements	30	2	4	⑪⑨
Arzu et al. 2024 [52]	Turkey	17/17	21.93 ± 1.90	Yoga / No arrangements	60	2	8	⑪⑬
Chang et al. 2023 [43]	Taiwan, China	65/63	32–36	Yoga / No arrangements	30	3	12	⑤
Ghafoureh et al. 2018 [38]	Iran	26/28	34.462 ± 5.368/30.179 ± 6.29	Yoga / Daily activities	60	3	8	⑥⑫
Ghafoureh et al. 2019 [49]	Iran	26/28	34.462 ± 5.368/30.179 ± 6.29	Yoga / Daily activities	60	3	10	⑮
Riddhi Sahua et al. 2022 [40]	India	30/30	38.37 ± 5.52	Yoga / No arrangements	45	7	10	⑭
Prem Lata et al. 2018 [46]	India	30/30	17–28	Yoga / No arrangements	45	7	10	⑭
Farzane et al. 2024 [36]	Iran	30/30	21.27 ± 2.5/21.63 ± 1.73	Running/ No arrangements	20 ~ 30	3	8	⑪⑬

T Experimental group, C Control group, T1 Experimental group 1, T2 Experimental group 2, Min minute, T/W Time/Week, W Week, / Not reported, ① The questionnaires of temporary determination of PMS, ② Recorded daily symptoms of PMS, ③ modified Premenstrual Syndrome Questionnaire, ④ a modified version of standard questionnaire for assessing symptoms of menstrual Disorders, ⑤ the daily record of severity of problem, ⑥ Beck Depression, ⑦ McGill Pain Questionnaire, ⑧ Perceived Stress Scale, ⑨ Visual Analogue Scale , ⑩ World Health Organization Quality of Life Assessment Scale, ⑪ Premenstrual Syndrome Scale, ⑫ Premenstrual Symptoms Screening Tool, ⑬ The Depression Anxiety Stress Scale-21, ⑭ Moos Menstrual Distress Questionnaire, ⑮ the Hamilton anxiety rating scale, ⑯ Borg scale

Exercise parameters across studies were as follows: intervention period = 4–12 weeks, intervention duration = 20–60 min, and frequency = 1–7 sessions per week.

Methodological quality

All included studies were RCTs. PEDro scores ranged from 4 to 7, with a mean of 6.2, indicating overall moderate-to-good quality. (Table 3)

Table 3.

Quality assessment of included studies using the PEDro scale

Included study	Random Allocation	Concealed allocation	Concealed allocation	Blinding of subjects	Blinding of therapists	Blinding of outcome assessment	Dropout rat ≤ 15%	Intention analysis	Between- group statistical comparisons	Point measures and measures of variability	Total
Hassan et al.2025 [50]	1	0	1	0	0	0	1	1	1	1	5
Zahra et al. 2018 [35]	1	0	1	0	0	0	1	1	1	1	6
A.El-Lithy et al.2014 [48]	1	0	1	0	0	0	1	1	1	1	6
Elvan et al. 2019 [51]	1	1	1	0	0	0	1	1	1	1	7
Mitra et al. 2012 [35]	1	0	1	0	0	0	1	1	1	1	6
Ayça et al. 2023 [41]	1	1	1	0	0	0	1	1	1	1	7
Elif et al. 2020 [39]	0	0	1	0	0	0	0	1	1	1	4
Şeyma et al. 2024 [45]	1	1	1	0	0	0	1	1	1	1	7
Meryem et al. 2024 [44]	1	0	1	0	0	0	1	1	1	1	6
Mahin et al. 2017 [47]	1	1	1	0	0	0	1	1	1	1	7
Didem et al. 2020 [37]	1	1	1	0	0	0	1	1	1	1	7
Duygu et al. 2023 [42]	1	1	1	0	0	0	1	1	1	1	7
Ece Karali et al. 2024 [34]	0	0	1	0	0	0	0	1	1	1	5
Arzu et al. 2024 [52]	1	1	1	0	0	0	1	1	1	1	7
Chang et al. 2023 [43]	1	0	1	0	0	0	0	1	1	1	5
Ghafoureh et al. 2018 [38]	1	1	1	0	0	0	1	1	1	1	7
Ghafoureh et al. 2019 [49]	1	1	1	0	0	0	1	1	1	1	7
Riddhi Sahua et al.2022 [40]	1	0	1	0	0	0	1	1	1	1	6
Prem Lata et al. 2018 [46]	1	0	0	0	0	1	1	1	1	1	6
Farzane et al. 2024 [36]	1	1	1	0	0	0	1	1	1	1	7

Meta-analysis

Effects of exercise on negative affect

Fourteen studies (n = 844) compared changes in negative affect between exercise and control groups. As shown in Fig. 2, heterogeneity was high (I² = 71.03%, P < 0.01), therefore, a random-effects model was applied. The pooled analysis demonstrated a significant reduction in negative affect following exercise (SMD = − 0.81, 95% CI [–0.96, − 0.66], P < 0.01), indicating that exercise improves mood disturbances in PMS patients.

Fig. 2.

Forest plot of exercise effects on negative affect in PMS. (① anxiety, ② depression, ③ Irritability, ④ Depressive Affect, ⑤ Negative effect, ⑥ Depressive thoughts, ⑦ emotional, ⑧ Depressive symptoms, ⑨ Depressed/sad/blue, ⑩ Anger/irritability, ⑪ Angry, irritable, ⑫ Depressed mood, ⑬ Anxious mood, T1 Experimental group 1, T2 Experimental group 2)

Effects of exercise on pain

Ten studies (n = 483) assessed pain outcomes. As shown in Fig. 3, substantial heterogeneity was observed (I² = 92.72%, P < 0.01), hence, a random-effects model was used. Exercise significantly reduced pain compared with control conditions (SMD = − 1.04, 95% CI [–1.67, − 0.41], P < 0.01).

Fig. 3.

Forest plot of exercise effects on pain symptoms in PMS. (① Premenstrual Syndrome Scale, ② Visual Analogue Scale)

Effects of exercise on fatigue

Ten studies (n = 497) evaluated fatigue outcomes. As shown in Fig. 4, heterogeneity was moderate (I² = 65.16%, P < 0.01), the random-effects model showed a significant reduction in fatigue following exercise (SMD = − 0.54, 95% CI [–0.84, − 0.23], P < 0.01).

Fig. 4.

Forest plot of exercise effects on fatigue in PMS

Subgroup analyses

To explore potential sources of heterogeneity, subgroup analyses were conducted by exercise intervention period (4–8 weeks vs. 9–12 weeks), intervention duration (20–40 min vs. 41–60 min), frequency (< 3 sessions/week vs. ≥ 3 sessions/week), and the type of exercise (running, Pilates, yoga).For negative affect, further subgroups were classified as depression, anxiety, and irritability.

As shown in Table 4, in the subgroup analysis for negative affect, running and Pilates interventions, an intervention period of 9–12 weeks, an intervention duration of 41–60 min, and a frequency of ≥ 3 sessions per week were associated with a large effect size (SMD: − 0.987, − 0.920, − 0.942, − 0.860, and − 0.878, respectively; SMD ≥ 0.8). In contrast, yoga, an intervention period of 4–8 weeks, an intervention duration of 20–40 min, and a frequency of < 3 sessions per week yielded a moderate effect (SMD: − 0.716, − 0.608, − 0.732, and − 0.585, respectively; 0.5 ≤ SMD < 0.8).In the subgroup analysis for pain, yoga, 9–12 weeks, 41–60 min, and ≥ 3 sessions per week were associated with a large effect size (SMD: − 1.025, − 1.507, − 1.030, and − 1.507, respectively; SMD ≥ 0.8). Interventions lasting 4–8 weeks and those with a frequency of < 3 sessions per week showed a small effect size (SMD: − 0.377 for both; 0.2 ≤ SMD < 0.5).Regarding fatigue, yoga was associated with a large effect size (SMD: − 0.842; SMD ≥ 0.8). 9–12 weeks, 20–40 min, and frequencies of both ≥ 3 and < 3 sessions per week showed a moderate effect size (SMD: − 0.627, − 0.644, − 0.503, and − 0.610, respectively; 0.5 ≤ SMD < 0.8), whereas 4–8 weeks and 41–60 min resulted in a small effect size (SMD: − 0.409 and − 0.436, respectively; 0.2 ≤ SMD < 0.5).

Table 4.

Subgroup analyses of exercise effects on negative affect, pain, and fatigue

Outcome	Adjustment Variables	Subgroup Category	Amount of study	N	SMD	95% CI	P	I2/%
Negative Affect	The type of exercise	Running	3	150	-0.987	[-1.247, -0.727]	< 0.01	38.79
		Pilates	2	110	-0.920	[-1.258, -0.583]	< 0.01	66.19
		Yoga	9	530	-0.716	[-0.922, -0.509]	< 0.01	76.59
	Period/week	4–8	5	288	-0.608	[-0.738, -0.478]	< 0.01	2.21
		9–12	9	556	-0.942	[-1.175, -0.710]	< 0.01	79.41
	intervention duration/min	20–40	4	298	-0.732	[-0.846, -0.618]	< 0.01	0
		41–60	10	492	-0.860	[-1.101, -0.618]	< 0.01	77.86
	Frequency (times/week)	≥ 3	11	616	-0.878	[-1.070, -0.686]	< 0.01	75.76
		<3	3	174	-0.585	[-0.745, -0.425]	< 0.01	0
	The type of Negative Affect	Depression	9	556	-0.752	[-0.933, -0.571]	< 0.01	50.04
		Anxiety	9	556	-0.760	[-0.983, -0.536]	< 0.01	48.65
		Irritability	6	432	-0.617	[-0.896, -0.338]	< 0.01	61.12
Pain	The type of exercise	Pilates	3	156	-0.652	[-1.438, 0.135]	0.104	82.98
		Yoga	4	257	-1.025	[-1.929,-0.121]	0.026	94.51
	Period/week	4 ~ 8	3	186	-0.377	[-0.620, -0.134]	0.002	0
		9 ~ 12	5	297	-1.507	[-2.410,0.603]	0.001	93.15
	intervention duration/min	20 ~ 40	2	147	-1.066	[-2.284, 0.151]	0.086	94.41
		41 ~ 60	6	336	-1.03	[-1.834, -0.225]	0.012	93.07
	Frequency (times/week)	≥ 3	5	186	-1.507	[-2.410,0.603]	0.001	93.15
		<3	3	297	-0.377	[-0.620,0.134]	0.002	0
Fatigue	The type of exercise	Pilates	2	110	-0.221	[-0.626,0.185]	0.286	16.22
		Yoga	4	322	-0.842	[-1.078,0.176]	0.000	11.89
	Period/week	4 ~ 8	3	205	-0.409	[-0.842,0.025]	0.065	58.8
		9 ~ 12	4	292	-0.627	[-1.078,0.176]	0.006	70.99
	intervention duration/min	20 ~ 40	3	273	-0.644	[-1.263, -0.025]	0.042	83.54
		41 ~ 60	4	224	-0.436	[-0.727,0.145]	0.003	18.46
	Frequency (times/week)	≥ 3	5	357	-0.503	[-0.93,-0.076]	0.021	74.31
		<3	2	140	-0.610	[-0.961,-0.259]	0.001	6.28

As shown in Table 4, Pilates interventions and 20–40-minute sessions showed no statistically significant effect on pain. Similarly, Pilates and 4–8-week interventions were non-significant for fatigue. All other subgroup effects were statistically significant.

Sensitivity analysis

Sensitivity analyses were performed to examine the influence of individual studies on overall heterogeneity.

For negative affect, the pooled estimate across all studies was SMD = − 0.81 (95% CI [–0.96, − 0.66], I² = 71.03%). When individual studies were sequentially removed, pooled SMDs ranged from − 0.83 to − 0.78, and I² varied between 63.07% and 72.00%, indicating stable results (P < 0.01). (Table 5 )

Table 5.

Sensitivity analysis after excluding individual studies

Outcome	excluding individual studies	SMD	95%CI	P	I2/%
Negative Affect	A.El-Lithy2014①	-0.79	[-0.94, -0.64]	< 0.01	70.27%
	A.El-Lithy2014②	-0.80	[-0.95, -0.65]	< 0.01	71.30%
	Mitra 2012(T1)	-0.80	[-0.95, -0.65]	< 0.01	70.96%
	Mitra 2012(T2)	-0.80	[-0.95, -0.65]	< 0.01	71.07%
	Ayça Balmumcu2023①	-0.82	[-0.98, -0.66]	< 0.01	71.72%
	Ayça Balmumcu2023④	-0.81	[-0.97, -0.66]	< 0.01	71.98%
	Ayça Balmumcu2023⑥	-0.83	[-0.98, -0.67]	< 0.01	70.96%
	Ayça Balmumcu2023③	-0.82	[-0.96, -0.67]	< 0.01	71.43%
	Elif2020①	-0.80	[-0.95, -0.64]	< 0.01	70.89%
	Elif2020④	-0.80	[-0.96, -0.65]	< 0.01	71.32%
	Elif2020⑥	-0.80	[-0.95, -0.64]	< 0.01	70.54%
	Elif2020③	-0.80	[-0.95, -0.64]	< 0.01	70.56%
	Mahin Kamalifard2017⑦	-0.78	[-0.93, -0.64]	< 0.01	67.31%
	Duygu Korkem Yorulmaz2023①	-0.82	[-0.98, -0.67]	< 0.01	71.34%
	Duygu Korkem Yorulmaz2023④	-0.81	[-0.97, -0.66]	< 0.01	71.99%
	Duygu Korkem Yorulmaz2023⑥	-0.82	[-0.98, -0.67]	< 0.01	71.44%
	Duygu Korkem Yorulmaz2023③	-0.82	[-0.98, -0.67]	< 0.01	71.59%
	Ece Karali2024①	-0.82	[-0.98, -0.67]	< 0.01	71.44%
	Ece Karali2024④	-0.82	[-0.98, -0.66]	< 0.01	71.71%
	Ece Karali2024⑥	-0.82	[-0.97, -0.66]	< 0.01	71.74%
	Ece Karali2024③	-0.82	[-0.98, -0.67]	< 0.01	71.08%
	Arzu Abic2023②	-0.80	[-0.95, -0.65]	< 0.01	71.23%
	Arzu Abic2023①	-0.80	[-0.96, -0.65]	< 0.01	71.52%
	Hsing-Chi Chang2023⑧	-0.81	[-0.97, -0.65]	< 0.01	71.59%
	Hsing-Chi Chang2023①	-0.81	[-0.96, -0.65]	< 0.01	71.34%
	Hsing-Chi Chang2023⑨	-0.82	[-0.98, -0.66]	< 0.01	71.25%
	Hsing-Chi Chang2023⑩	-0.81	[-0.97, -0.66]	< 0.01	71.61%
	Hsing-Chi Chang2023⑪	-0.81	[-0.97, -0.66]	< 0.01	71.61%
	Ghafoureh Ghaffarilaleh2018②	-0.81	[-0.96, -0.65]	< 0.01	71.74%
	Ghafoureh Ghaffarilaleh2018③	-0.83	[-0.98, -0.68]	< 0.01	69.96%
	Ghafoureh Ghaffarilaleh2019①	-0.82	[-0.97, -0.66]	< 0.01	71.89%
	Ghafoureh Ghaffarilaleh2019⑫	-0.83	[-0.98, -0.68]	< 0.01	69.06%
	Ghafoureh Ghaffarilaleh2019⑬	-0.82	[-0.98, -0.67]	< 0.01	71.27%
	Riddhi Sahu2022⑤	-0.83	[-0.98, -0.68]	< 0.01	70.03%
	Prem Lata2018⑤	-0.77	[-0.91, -0.63]	< 0.01	63.07%
	Farzane Shavaisi12024(T1)①	-0.82	[-0.98, -0.66]	< 0.01	71.74%
	Farzane Shavaisi12024(T2)①	-0.82	[-0.97, -0.66]	< 0.01	71.92%
	Farzane Shavaisi12024(T1)②	-0.81	[-0.97, -0.66]	< 0.01	72.00%
	Farzane Shavaisi12024(T2)②	-0.81	[-0.97, -0.66]	< 0.01	71.99%
Pain	Elvan YILMAZ-AKYUZ2019	-0.90	[-1.52, -0.27]	< 0.01	92.02%
	Ayça Balmumcu2023	-1.13	[-1.81, -0.44]	< 0.01	92.99%
	Elif2020	-0.99	[-1.69, -0.30]	< 0.01	93.57%
	Şeyma Aykut2024	-1.13	[-1.81, -0.45]	< 0.01	93.14%
	Duygu Korkem Yorulmaz2023⑭	-1.09	[-1.79, -0.38]	< 0.01	93.40%
	Duygu Korkem Yorulmaz2023⑮	-1.07	[-1.77, -0.36]	< 0.01	93.52%
	Ece Karali2024⑭	-1.08	[-1.79, -0.58]	< 0.01	93.25%
	Ece Karali2024⑮	-1.13	[-1.81, -0.45]	< 0.01	92.73%
	Riddhi Sahu2022	-1.10	[-1.80, -0.41]	< 0.01	93.28%
	Prem Lata2018	-0.78	[-1.21, -0.35]	< 0.01	83.90%
Fatigue	Zahra Mohebbi Dehnavi2018	-0.63	[-0.93, -0.33]	< 0.01	57.54%
	Ayça Balmumcu2023	-0.55	[-0.91, -0.19]	< 0.01	70.72%
	Elif2020	-0.62	[-0.93, -0.31]	< 0.01	60.45%
	Duygu Korkem Yorulmaz2023	-0.52	[-0.88, -0.16]	< 0.01	70.71%
	Ece Karali2024	-0.49	[-0.85, -0.14]	< 0.01	68.61%
	Ghafoureh Ghaffarilaleh2018	-0.51	[-0.86, -0.15]	< 0.01	70.47%
	Hsing-Chi Chang2023	-0.43	[-0.70, -0.15]	< 0.01	42.69%

① anxiety, ② depression, ③ Irritability, ④ Depressive Affect, ⑤ Negative effect, ⑥ Depressive thoughts, ⑦ emotional, ⑧ Depressive symptoms, ⑨ Depressed/sad/blue, ⑩ Anger/irritability, ⑪ Angry/irritable, ⑫ Depressed mood, ⑬ Anxious mood, T1 Experimental group 1, T2 Experimental group 2, ⑭ Premenstrual Syndrome Scale, ⑮ Visual Analogue Scale

For pain, the overall effect was SMD = − 1.04 (95% CI [–1.67, − 0.41], I² = 92.72%), removing single studies yielded SMDs between − 1.13 and − 0.78, with I² = 83.90–93.52%, suggesting robust findings (P < 0.01). (Table 5 )

For fatigue, the pooled effect was SMD = − 0.54 (95% CI [–0.84, − 0.23], I² = 65.16%). Excluding the study by Hsing-Chi Chang et al. [43] markedly reduced heterogeneity (SMD = − 0.43, 95% CI [–0.70, − 0.15], I² = 42.69%), indicating that differences in group allocation methods—participants from two distinct regions assigned as experimental vs. control groups—may explain part of the observed heterogeneity. (Table 5 )

Publication bias

Egger’s test revealed no significant publication bias for fatigue (|t| = 2.1, P > 0.05), but significant bias for negative affect (|t| = − 3.86, P < 0.05) and pain (|t| = − 5.35, P < 0.05).Application of the non-parametric trim-and-fill method showed that the pooled estimates and 95% CIs remained stable after adjustment. This suggests that while funnel plot asymmetry was observed, it does not undermine the overall findings, and the results remain relatively robust. (Fig. 5)

Fig. 5.

Funnel plots for publication bias (from left to right: negative affect, pain, and fatigue)

Quality of evidence

GRADE assessment indicated moderate evidence quality for fatigue and low evidence quality for negative affect and pain (Table 6).Downgrades were primarily due to incomplete blinding or allocation concealment in several studies and the presence of publication bias in two outcome domains.

Table 6.

GRADE summary of evidence quality

Outcome	Included study	Evidence Quality Assessment					Evidence Quality
		Limitations	Heterogeneity	Indirectness	Imprecision	Publication bias
Negative Affect	14	Non-serious	Serious	Non-serious	Non-serious	Serious	Low
Pain	10	Non-serious	Serious	Non-serious	Non-serious	Serious	Low
Fatigue	10	Non-serious	Serious	Non-serious	Non-serious	Non-serious	Moderate

Discussion

This meta-analysis demonstrated that exercise alleviates negative affect, pain, and fatigue in women with premenstrual syndrome (PMS), which aligns in part with findings from previous studies [12, 20–22, 54, 55]. Two meta-analyses focusing on yoga interventions confirmed its efficacy for PMS symptom relief but were limited to a single exercise modality [20, 21]. Another meta-analysis on exercise interventions in PMS summarized overall benefits but included too few RCTs to permit subgroup analyses [22]. Other systematic reviews likewise supported the therapeutic role of exercise for PMS symptom management [12, 54, 55], consistent with the present findings. However, one study reported that Pilates had no significant effect on fatigue [39]—a result inconsistent with our pooled estimates. This discrepancy may be attributed to the intervention design: the non-significant study used only basic Pilates movements without progression, limiting stimulus intensity and adaptation potential for addressing fatigue. By contrast, another trial [41] implemented a three-phase progressive program (beginner–intermediate–advanced), enhancing both exercise intensity and physiological adaptability, thereby achieving greater improvements in fatigue. Physiological evidence further supports these findings. During the luteal phase, elevated plasma levels of estrogen and progesterone have been associated with increased severity of premenstrual symptoms [56], whereas aerobic exercise has been shown to reduce these hormonal levels while relieving PMS symptoms [48].

The methodological quality of included trials, assessed via the PEDro scale, averaged 6.2 points, indicating moderate quality overall. Nonetheless, most studies lacked allocation concealment and blinding of participants, therapists, or assessors, introducing potential bias. The absence of stratification by PMS severity and the high heterogeneity (I² > 50%) also suggest variability across studies. Still, the apparent dose–response relationship between exercise parameters and symptom improvement strengthened the certainty of evidence, warranting an upward adjustment in GRADE ratings. Accordingly, evidence quality was graded low for negative affect and pain, and moderate for fatigue outcomes. Furthermore, in analyzing the baseline characteristics of the included studies, we observed a notably clustered geographical distribution, with the majority of studies originating from regions such as Turkey and Iran. This distribution pattern may be associated with the higher reported prevalence of PMS in these areas. For instance, one study indicated that PMS is highly prevalent among reproductive-aged women in Turkey, with a reported rate of 52.2% [57].

Our subgroup results indicate that running for 9–12 weeks at a frequency ≥ 3 sessions per week appeared to be more effective in negative affect. Exercise stimulates the release of neurotransmitters that promote positive emotional states [58]. After six weeks of jogging, plasma γ-aminobutyric acid (GABA) levels significantly increase, accompanied by reduced depressive symptoms [59]. Jiayu et al. [60]reported that a 12-week intervention produced greater improvements in depression than a 6-week program. Similarly, eight weeks of voluntary running elevated dopamine and its metabolites in the prefrontal cortex and striatum of rats [61]. Zhuo et al. [62]found that women participating in physical education fewer than twice per week had a 4.43-fold higher risk of developing PMS compared with those exercising ≥ 2 times per week. Moreover, high-frequency (3–5 sessions/week) aerobic exercise over eight weeks increased serotonin in the hippocampus and norepinephrine in the amygdala of patients with depression [63], while frequent training accelerated serum BDNF peaks and enhanced central neurotransmitter adaptation, facilitating greater mood stabilization [64]. Dysregulation of the hypothalamic–pituitary–adrenal (HPA) axis has long been linked to anxiety and depression [65], preclinical research shows that voluntary exercise modulates hypothalamic CRF and pituitary ACTH release [66], thereby improving stress responsiveness and anxiety regulation [67].

For pain relief, our analysis suggests that yoga practiced for 9–12 weeks at ≥ 3 sessions per week appeared to be more effective. Yoga may modulate pain perception and tolerance by regulating stress reactivity [68, 69]. Short-term yoga has been reported to enhance attention and well-being in women with PMS, possibly through increased alpha-wave activity and slower diaphragmatic breathing [70]. Enhanced alpha-band EEG activity correlates with relaxation, emotional stability, and increased serotonin release, collectively improving subjective well-being after yoga practice [71]. Yi et al. [72]observed that exercise lasting ≥ 8 weeks and performed more than three times weekly effectively reduced visual analogue scale (VAS) pain scores in primary dysmenorrhea. Another study [73] showed that performing 45–60 min of exercise at least three times weekly produced an average 25 mm reduction in pain intensity on a 100 mm VAS—considered clinically meaningful. The analgesic mechanism of exercise may involve the release of β-endorphins, which inhibit nociceptive pathways and elevate pain thresholds [74, 75].

Regarding fatigue, the most effective regimen also appeared to be yoga performed for 9–12 weeks at ≥ 3 sessions per week. Yoga alleviates fatigue by strengthening musculature, enhancing flexibility, improving respiratory efficiency, and regulating visceral, vascular, and neural circulation, thereby inducing relaxation and comfort [47]. However, evidence remains limited concerning the optimal duration and frequency of exercise for fatigue reduction, underscoring the need for further large-scale, high-quality trials.

Limitations and future perspectives

1.  Although subgroup analyses explored exercise type, duration, frequency, and period, the optimal parameters for each outcome remain uncertain, requiring additional studies for confirmation.

2.  Heterogeneity may arise from variations across the included studies. For instance, one study exclusively enrolled participants with moderate-to-severe PMS [46], excluding those with mild or severe symptoms. Differences in inclusion thresholds were noted, with some studies applying a PMSS score ≥ 110 [37, 52]while others used a cutoff > 88 [39]. Measurement tools also varied: several studies employed the Premenstrual Syndrome Scale, which covers multiple subdomains including depression, anxiety, pain, and fatigue [34, 36, 37, 39, 41, 42, 45, 51, 52], whereas others used domain‑specific instruments for negative affect (e.g., Hamilton Anxiety Rating Scale, Beck Depression Inventory) [35, 38, 49], pain (e.g., Visual Analogue Scale) [34, 42, 44, 51], or fatigue (e.g., Perceived Stress Scale) [44, 51]. Methodological diversity was further observed in randomization, with one trial using cluster randomization [43]. Collectively, these discrepancies in population selection, outcome assessment, and study design likely contributed to the substantial heterogeneity observed, which may affect the reliability of the pooled estimates.

3. Egger’s test indicated significant publication bias for certain outcomes. Although the trim‑and‑fill method was applied to adjust for asymmetry, residual bias cannot be fully ruled out. Further high‑quality, well‑designed studies are needed to strengthen the evidence base.

4. The range of exercise modalities included in this meta‑analysis remains limited. Future research should investigate a broader variety of exercise types to better inform practical recommendations.

Conclusion

Exercise alleviates negative affect, pain, and fatigue in women with premenstrual syndrome (PMS). For improving negative affect, running performed for 20–40 min per session at a frequency of at least three times per week is recommended. To relieve pain, yoga practiced three or more times per week for 9–12 weeks appears most beneficial. Similarly, yoga undertaken three or more times weekly for 9–12 weeks is optimal for mitigating fatigue. These findings provide practical guidance for the individualized design of exercise prescriptions in PMS management.

Authors’ contributions

X.Y.L. , X.C.M , F.D. , J.W.C. , X.W. , L.J.C. conceptualized the design and drafted the initial manuscript. F.D. and X.C.M collected data and conducted data analysis. L.J.C. performed the final review of the manuscript. All authors have read and agreed to the published version of this manuscript.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.