Exercise for premenstrual syndrome: a systematic review and meta-analysis of randomised controlled trials

Background Exercise is recommended as a treatment for premenstrual syndrome (PMS) in clinical guidelines, but this is currently based on poor-quality trial evidence. Aim To systematically review the evidence for the effectiveness of exercise as a treatment for PMS. Design & setting This systematic review searched eight major databases, including MEDLINE, EMBASE and the Cochrane Central Register of Controlled Trials (CENTRAL), and two trial registries from inception until April 2019. Method Randomised controlled trials (RCTs) comparing exercise interventions of a minimum of 8-weeks duration with non-exercise comparator groups in women with PMS were included. Mean change scores for any continuous PMS outcome measure were extracted from eligible trials and standardised mean differences (SMDs) were calculated where possible. Random-effects meta-analysis of the effect of exercise on global PMS symptoms was the primary outcome. Secondary analyses examined the effects of exercise on predetermined clusters of psychological, physical, and behavioural symptoms. Results A total of 436 non-duplicate returns were screened, with 15 RCTs eligible for inclusion (n = 717). Seven trials contributed data to the primary outcome meta-analysis (n = 265); participants randomised to an exercise intervention reported reduced global PMS symptom scores (SMD = -1.08; 95% confidence interval [CI] = -1.88 to -0.29) versus comparator, but with substantial heterogeneity (I 2 = 87%). Secondary results for psychological (SMD = -1.67; 95% CI = -2.38 to -0.96), physical (SMD = -1.62; 95% CI = -2.41 to -0.83) and behavioural (SMD = -1.94; 95% CI = -2.45 to -1.44) symptom groupings displayed similar findings. Most trials (87%) were considered at high risk of bias. Conclusion Based on current evidence, exercise may be an effective treatment for PMS, but some uncertainty remains.


Introduction
PMS describes a broad collection of psychological, physical, and behavioural symptoms, occurring in line with the menstrual cycle. [1][2][3] Up to 75% of women may experience symptoms of PMS during their lifetime. 1,4 Women with PMS frequently present to primary care with acute symptoms, as well as decreased work productivity and relationship difficulties; PMS can also be associated with conditions such as hypertension and depression. [5][6][7] Many therapies are recommended for symptom management; for example, both the National Institute for Health and Care Excellence (NICE) and the Royal College of Obstetricians and Gynaecologists (RCOG) suggest exercise as a first-line treatment, alongside medications such as selective serotonin reuptake inhibitors and the combined oral contraceptive pill. 1,8 The latter two treatments are often effective, but risk side effects such as fatigue, nausea and preclusion of, or potential risks to, pregnancy. 9,10 Women may also actively prefer alternative therapies. 11 Nevertheless, exercise is known to increase endorphin levels, to help regulate progesterone and oestrogen synthesis and to encourage the production of endogenous anti-inflammatory chemicals. [12][13][14] Exercising also brings other benefits such as improved overall fitness, opportunities to socialise, and the potential for reduction in feelings of depression, all of which may help to moderate the symptom profile in PMS. 2,15,16 A previous systematic review of this topic found low-level evidence to support the recommendation of exercise for symptom management in PMS. 17 It is this review that provides the supporting basis for the RCOG recommendation in its PMS guidelines.

Study aim
This study aimed to provide an updated systematic review and meta-analysis of the evidence examining the effectiveness of exercise as a treatment for PMS. It also undertook to examine the global effect of exercise on PMS symptoms, and the specific effects on each of three symptom domains: psychological, physical, and behavioural. These domains are those specified in the diagnostic questionnaire recommended for use in the UK, the Daily Record of Severity of Problems. 18

Method Data sources and searches
Cochrane Collaboration guidance on systematic reviews of interventions and PRISMA reporting guidelines were followed throughout. 19,20 The following databases were searched electronically for eligible trials: MEDLINE, EMBASE, PSYCInfo, CINAHL, AMED, SPORTDiscus, CENTRAL, and clinicaltrials. gov (MEDLINE strategy available at Supplementary figure S1). The World Health Organization International Clinical Trials Registry Platform was searched for trials in progress. The Directory of Open Access Journals was searched; grey literature was sought from OpenGrey; and the authors undertook forward and backward citation searching of included trials.
MeSH and free-text terms relating to the condition of interest (premenstrual syndrome, PMS, premenstrual dysphoric disorder, PMDD, and premenstrual tension) were combined with intervention terms (sport, exercise, physical activity). Scottish Intercollegiate Guideline Network study design limits were applied where possible. 21 No restrictions were placed by date, language, or country of origin. Searches were undertaken from individual database inception up to April 2018 and refreshed in April 2019; one new article was found at this point, however, the data were already included in the meta-analysis. The authors were contacted where further information was required. Two researchers independently screened all titles and abstracts before reviewing all full-text articles. Any discrepancies were discussed with a third researcher.

Study selection
Inclusion criteria used were: (a) populations of women reporting regular menstrual cycles (21-35 days) and symptoms of PMS; (b) trials delivering any exercise intervention for a minimum of 8 weeks, including co-interventions such as nutritional supplements, if delivered to both the exercise and nonexercise groups; (c) a non-exercise comparator; (d) any outcome measure quantifying PMS symptoms on a continuous scale; and (e) RCTs. Studies comparing two types of exercise without a non-exercise comparator were excluded.

Data extraction and quality assessment
Data were independently extracted by two reviewers using an a priori designed and piloted data extraction form. Data from referenced study protocols and author correspondence (where applicable and available) were also used to provide additional information to undertake the quality assessment.
Study quality was independently assessed by three reviewers using quality criteria from the Cochrane Risk of Bias Tool and adapted for inability to blind to group allocation in exercise trials. Criteria concerning selection, performance, detection, attrition, and reporting biases were included and were given equal weighting in the final assessment of overall trial quality. Additional items relating to recall and observer bias, the Hawthorne effect, and contamination bias were also assessed as a group to form the final part of the study quality score.
The quality of the evidence included was assessed by two reviewers using the GRADE (Grading of Recommendations Assessment, Development and Evaluation) methodology for the primary outcome, and all three secondary outcomes; the checklist by Meader et al was used to guide this assessment. 22 Data synthesis and analysis SMD and the associated 95% CIs were calculated using standard methods provided by the Cochrane Handbook for Systematic Reviews, and guided by a previous systematic review on a related topic. 16,23 Where studies reported >1 intervention arm testing similar exercise interventions with differing intensities but only one control group, intervention scores were combined to produce one group as per Cochrane Collaboration guidance. 24 Random-effects meta-analysis was used as heterogeneity within the results was expected.
Between-studies heterogeneity was examined using the I 2 statistic, with cut-off points of 25%, 50%, and 75% considered for low, medium, and high heterogeneity, respectively. 25 Publication bias was assessed narratively by the review team; funnel plots were not constructed as <10 trials contributed data to either the primary or secondary outcome meta-analyses.
All statistical analyses were performed using RevMan 5.3 and IBM SPSS Statistics (version 24). Following removal of duplicates, 435 titles and abstracts were screened. Of these, 58 full texts were reviewed and 20 reports of 15 individual trials were found to be eligible for inclusion. A further published report of the same trial was found on the PubMed search carried out in April 2019, giving 21 reports of 15 trials in total ( Figure 1 and Supplementary Table S1). 26

Risk of bias in included studies and publication bias
Overall, quality of the included trials was low; two 31,39 were considered unclear (moderate) risk of bias, and the remainder were considered high risk ( Figure 2). Identified issues that introduced risk of bias included lack of robust procedures for randomisation, minimal information regarding blinding of participants or assessors, and lack of pre-specified outcomes. Overall assessment of the evidence concluded that there may be a moderate indication of publication bias owing to a lack of negatively reporting trials.

Study results
All 15 of the included trials reported that exercise significantly decreased symptoms of PMS. Fourteen of the 15 trials reported their data in a format that could be used in meta-analysis of either primary or secondary outcome; Ilka et al did not report any standard deviations associated with mean scores or provide adequate data to calculate them, and the author could not be contacted. 29 The trial conducted by Pazoki et al contained four groups, analysed in this review as two pairs: a comparison of exercise intervention versus control, and a second comparison of exercise intervention plus fennel extract versus fennel extract alone. 34

secondary outcomes
Eight of the 15 trials reported data exploring the effect of exercise on psychological symptoms (n = 427 participants). 27,28,30,31,35,[38][39][40] The meta-analysis showed a statistically significant reduction in SMD for psychological symptoms in the exercise group relative to comparators (SMD = -1.67; 95% CI = −2.38 to -0.96). Significant heterogeneity was present (I 2 = 89%). Seven of the 15 trials reported data exploring the effect of exercise interventions on physical symptoms of PMS (n = 397 participants), 27,30,31,35,[38][39][40] and meta-analysis showed a statistically significant reduction in SMD for women randomised to exercise interventions compared with comparators (SMD = -1.62; 95% CI = Figure 2 Forest plot for primary outcome: exercise versus control for overall symptom reduction. -2.41 to -0.83). Heterogeneity was significant (I 2 = 91%). Two of the 15 trials examined the effect of the exercise interventions on behavioural symptoms (n = 92 participants), 28,40 and meta-analysis found a statistically significant reduction in SMD relative to comparators (SMD = -1.94; 95% CI = -2.45 to -1.44) with I 2 = 0 (see Supplementary Figures S2-S4). GRADE assessment showed the overall strength of the evidence for all four outcomes to be moderate ( Figure 4). Evidence was downgraded from high strength owing to the large number of included trials at high or very high risk of bias.

Discussion Summary
This systematic review provides some further support for clinical guidelines that recommend exercise as an effective treatment for PMS. The secondary analyses conducted also provide new evidence that exercise might be useful in alleviating specific psychological, physical, and behavioural symptoms associated with PMS, as well as assisting with management of the global symptom profile.

Strengths and limitations
These results should be interpreted cautiously owning to the heterogeneity seen throughout the review. A wide variety of exercise types were included in this review, and the interventions also differed in length and setting. A random-effects meta-analysis model was used to account for this heterogeneity and subgroup analyses were carried out where possible. The initial intention was to further explore any heterogeneity in intervention type; however, too few studies were identified to enable this to be feasible. No studies followed-up women post-intervention so any lasting effects of exercise on symptoms cannot be assessed at this point. Twelve of the 15 included trials were conducted in Iran, where although the prevalence of PMS is similar to that of the UK, the cultural meaning around women's health and health care more generally may make these results less applicable for some UK women. Most studies recruited small samples, and many did not include a clear description of the interventions assessed, making classifications difficult. Only one of the 15 included trials clearly prespecified their outcomes, 36 and different outcome measures were used to assess symptoms across trials.
This review has several strengths. PRISMA guidelines were followed throughout, and the a priori registration of a protocol in PROSPERO utilises all current best-practice guidance for systematic reviews. A comprehensive search strategy, developed in line with similar reviews, was used and multiple electronic databases and grey literature were searched for eligible trials. No language restrictions were made; therefore, it is unlikely that eligible studies were missed for this reason. A native Persian speaker translated all Persian language studies. In only including RCTs in this review, it is hoped the highest quality of evidence currently available on this question has been utilised. The included trials represent women of a range of ages recruited from a variety of clinical and non-clinical settings. For the first time this review descriptively and quantitatively summarises the available evidence on the question of whether exercise is an effective treatment of the symptoms of PMS.

Comparison with existing literature
The results of this review are in overall agreement with an earlier systematic review by Daley, although that review used broader inclusion criteria, particularly in relation to study design. 17 The narrative aspect of this updated systematic review examined 15 trials that span a wider range of exercise types and included more participants than the previous review (n = 72 versus n = 717), but concerns about trial quality remain, along with insufficient reporting detail. The meta-analysis results reported here are consistent with other reviews that have reported exercise to be an effective intervention for primary dysmenorrhoea, and for improving mental and physical health outcomes in the population. 15,42,43 Exercise is also recommended by the American College of Sports Medicine and NICE to enhance the initial management and advice given on other lifestyle issues known to be associated with PMS, such as obesity and depression. [44][45][46] Implications for research and practice The results of this systematic review support the recommendation that GPs and other primary healthcare professionals caring for women with PMS should consider exercise as part of their treatment plan. The research recommendations are centred on improving reporting practices, and measures to improve the quality of future trials. It is already strongly suggested that future research in this area should use either the Daily Rating of Severity of Problems 18 or the Premenstrual Symptoms Screening Tool 47 alone,: both measure and report global symptom change, as well as symptom cluster results, such as those reported here. By re-administering one of these questionnaires at intervals of 6 months or 1 year, further information relating to the persistence of the positive effects of exercise beyond the end of the trial period could be investigated. If larger sample sizes are used, then subgroup analysis by age, contraceptive use, or other demographic factors may be carried out by future systematic reviewers. Finally, by using the internationally recognised Consolidated Standards of Reporting Trials guidelines, 48 then the changes suggested above will be clearly communicated to future clinicians and researchers, and can be acted on appropriately.
This systematic review finds that exercise may be an effective treatment for PMS. GPs and other primary healthcare professionals may wish to advise patients that exercise might help reduce their PMS symptoms, but should do so with caution until better-quality evidence becomes available.

Funding
No funding was received to support the undertaking of this study or the preparation of this manuscript.

Trial registration number
The review was prospectively registered on PROSPERO: CRD42018088418.

Provenance
Freely submitted; externally peer reviewed.