Effects of five types of exercise on vascular function in postmenopausal women: a network meta-analysis and systematic review of 32 randomized controlled trials

Registration

The literature search date for this study was January 1, 2000–December 31, 2023, and our actual literature search for this study was January 5, 2024.

The research protocol for this NMA was registered in the International Prospective Register of Systematic Reviews (identifier: CRD42024498890) and strictly adheres to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines for NMA (Hutton et al., 2015).

Literature search strategy

We conducted a systematic search in several databases including PubMed, Web of Science, Cochrane, EBSCO, and Embase. We applied search terms using MeSH terms in PubMed/Cochrane and Emtree terms in Embase. The primary terms used to retrieve relevant literature included exercise, training, aerobic exercise, moderate intensity continuous training, resistance training, combined training, and high-intensity interval training, combined with postmenopausal, postmenopausal*, older adult women, vascular function, endothelium, pulse wave velocity, and flow-mediated dilation. The search strategy was based on key phrases relevant to the PICOS (Population, Intervention, Comparison, Outcome, Study Design) tool: (P) Population: postmenopausal women and/or older adult women; (I) Intervention: CET, INT, RT, CT, HYB, and CON (as listed in Table 1); (C) Comparison: no exercise; (O) Outcome: FMD, PWV, AIx, IMT, NO; (S) Study type: RCTs. Please refer to Appendix S2 for details of the search strategy.

Eligibility criteria

The study was deemed eligible if it satisfied the following criteria: (1) it involved postmenopausal or older female participants; (2) it included interventions such as CET, INT, RT, CT, or HYB (as listed in Table 1); (3) it evaluated at least one vascular function indicator: FMD, PWV, AIx, IMT, or NO; (4) the experimental group underwent structured exercise training for ≥6 weeks; (5) it encompassed all English-language RCTs published from 2000 to December 2023.

Studies were excluded if they: (1) were duplicates, literature reviews, editorials, letters to editors, conference abstracts, or animal studies; (2) full text was unavailable or complete study data was difficult to obtain; (3) did not report on the vascular function indicators of interest for this review; (4) involved mixed interventions with exercise and other measures, e.g., exercise combined with medication, diet, etc.

Study selection

Title and abstract screening, as well as full-text screening, were conducted independently by two authors (Weihao Sun, Yanli Han) in accordance with the inclusion and exclusion criteria. Any disputes were resolved by Song Gu. The references of the selected articles were manually searched to identify articles that met the inclusion criteria.

Data extraction

Data from the included studies were independently extracted by two authors, Weihao Sun and Yanli Han. Any disagreement was resolved by the opinion of Song Gu. The literature provided the following information: the first author, publication year, country, and demographic characteristics of the participants (including the number of participants in the experimental and control groups, age, and comorbidities), details of the exercise intervention (such as the type of exercise, intensity, duration, frequency, session length, and whether it was supervised or unsupervised), and relevant outcome measures. If data was unavailable in the articles, we attempted to contact the authors to obtain missing information. Please refer to for further details.

Risk of bias assessment

The risk of bias (ROB) of the included studies was assessed by two researchers, Weihao Sun and Yanli Han, using the Cochrane Risk of Bias Assessment Tool (Nasser, 2020). Due to the nature of exercise interventions, blinding of participants was not feasible in the included studies (Huang et al., 2021). Therefore, we only assessed the other six domains of bias, which include random sequence generation, allocation concealment, blinding of outcome assessment, incomplete outcome data, selective reporting, and other sources of bias. We assessed the overall ROB for each study as follows: (1) Low ROB: studies not rated as having high ROB and with ≤3 domains judged as unclear risk (Chen et al., 2024); (2) Moderate ROB: studies with ≤1 domain rated as high ROB, but with ≥4 domains judged as unclear risk; (3) High ROB: all other scenarios categorized as high ROB (Cipriani et al., 2018).

Data synthesis and statistical analysis

This NMA estimated the effect by combining the pre-to-post changes in the experimental and control groups. SMDs and their 95% confidence intervals (Cis) for FMD, PWV, AIx, IMT, and NO were calculated. Random-effects multivariate NMA was conducted within a frequency framework (Salanti, 2012; Bucher et al., 1997) using STATA 17.0 software (StataCorp, College Station, TX, USA), following the current PRISMA NMA guidelines (Hutton et al., 2015). Estimates of combined effects, along with their 95% CIs and 95% prediction intervals (PrI), are presented.

A network diagram is generated to visually display the relationships between different intervention measures. The lines connecting the nodes represent direct comparisons between intervention measures, with the size of each node and the thickness of the connecting lines proportional to the number of studies. In addition, a network contribution plot is created to calculate the contributions of each direct comparison. The inclusion criteria of each study are assessed to determine whether postmenopausal women can be randomly assigned to receive any intervention. Consistency models are then used to evaluate the assumption of transitivity (Rücker & Schwarzer, 2015). Transitivity is a fundamental assumption of NMA. NMA assumes that indirect comparisons offer valid estimates of unobserved direct comparisons (Salanti, 2012). It also assumes that the effect modifiers of all studies follow a uniform distribution (Jansen & Naci, 2013). To assess the consistency of each closed loop, we calculate inconsistency factors (IFs) and their 95% CIs. We check for inconsistency using inconsistency models. When there is no significant inconsistency, a consistency model is adopted (P > 0.05) (Shim et al., 2017). Node-splitting analysis is conducted to examine local inconsistencies, ensuring reliable results (P > 0.05).

The surface under the cumulative ranking curve (SUCRA) was used to rank and compare the effects of different types of exercise interventions (Salanti, Ades & Ioannidis, 2011). SUCRA values range from 0 to 100, with higher values indicating better effects of exercise intervention (Mbuagbaw et al., 2017). Furthermore, a network funnel plot is constructed to test for the presence of publication bias in NMA, and a symmetry criterion is used for intuitive examination.

Statistical processing

Statistical processing data were analyzed by META through Stata17.0 software, with outcome indicators as continuous variables. Due to the varying use of tools for the same indicator in the included literature, the standardized mean difference (SMD) was used to indicate the baseline, and the baseline was uniformly adjusted to α = 0.05. Heterogeneity was tested using the Q-test and I2-test, and publication bias analysis was performed using the Egger’s test. In reticulated meta-analysis, if there was a closed-loop structure in the reticulated evidence graph, the inconsistency test was performed using nodal analysis. The results of the ring inconsistency test were analyzed for consistency modeling if P > 0.05. Local inconsistency tests were also conducted using node splitting, and if there was a result of P < 0.05, the results were directly compared in the conventional META analysis following this comparison. The results were analyzed by the SUCRA value of the area under the cumulative ranking probability map (surface under the cumulative ranking, SUCRA) to rank the superiority of each intervention (0 ≤ SUCRA ≤ 1).

Literature selection

Figure 1 depicts the flowchart of the literature search and screening process. The search strategy identified 645 articles, 643 from databases and two from reference lists. After removing 179 duplicates, 466 articles remained for screening. Of these, 298 were excluded after screening titles and abstracts. Upon full-text review, 136 articles were further excluded, leaving 32 articles eligible for quantitative synthesis.

Literature characteristics

The characteristics of the studies are outlined in Appendix S3, and the list of studies is provided in Appendix S4. The studies were published between 2000 and December 2023. The experimental group comprised 866 participants (CET: 341; RT: 235; INT: 83; CT: 68; HYB: 139), while the control group consisted of 561 participants aged between 50 and 70 years. Fourteen studies included participants with symptoms of obesity, hypertension, and cardiovascular risk. One study (Kujawski et al., 2018) included two males in each of the control group (n = 28) and experimental group (n = 27). Several included studies did not report the intensity of the HYB group, and the intensity of INT studies varied widely (e.g., 55–85% HRmax). The duration of exercise listed in Appendix S3 does not include warm-up, cool-down, or stretching time. Exercise interventions lasted for ≥6 weeks, with over half (78.12%) of the studies conducting interventions for 12 weeks. Twenty-three studies provided supervised training, six studies combined supervised and remote training, and four studies used remote training alone. The literature analyzed in this study mainly focused on FMD, PWV, AIx, IMT, and NO. Therefore, this NMA analyzed FMD, PWV, AIx, IMT, and NO. Although other vascular biomarkers, such as Endothelin-1 (ET-1) and β-stiffness index, were also studied, the data on these vascular biomarkers were insufficient to conduct a NMA.

Risk of bias assessment results

The risk of bias assessment results for each study are detailed in Appendix S5. Of the included studies, 30 reported clear randomization methods, 27 reported allocation concealment, and 23 described blinding in outcome assessment. A dropout rate exceeding 20% was considered high risk, and three studies were classified as high risk due to this criterion (Chen et al., 2024). Furthermore, 26 studies showed low risk of selective reporting. Among other biases, two studies were rated as high risk due to significant measurement errors in outcome assessment. In summary, 24 studies were judged to have low ROB, six had moderate ROB, and two had high ROB.

Network meta-analysis

Figure 2 shows the network diagram of eligible studies investigating the effectiveness of different exercise modalities on all five vascular function indicators. The size of the nodes corresponds to the sample size of each exercise modality, while the thickness of the lines between exercise modalities reflects the number of studies involved in that comparison. The most common intervention type is CET, while INT and CT have the lowest frequency. The contribution plot for network comparisons, both direct and indirect, and the number of studies for each direct comparison can be found in Appendix S6. Figure 3 shows forest plots of eligible comparisons of FMD, PWV, AIx, IMT, NO.

The study evaluated the inconsistency of vascular function outcome indicators through heterogeneity estimates, inconsistency models, and node-splitting analyses (refer to Appendix S7). Heterogeneity estimates for each loop showed non-significant consistency in FMD, PWV, IMT, NO, and AIx (P > 0.05), allowing for further comparisons. The inconsistency model demonstrated non-significance in both direct and indirect comparisons (P > 0.05). Node-splitting analysis showed no significant difference in both direct and indirect evidence (P > 0.05), indicating reliable results. Appendix S8 presents forest plots comparing qualified outcomes of vascular function indicators, including 95% CI and 95% PrI. Appendix S9 provides funnel plots for vascular function outcome indicators, which were used to examine NMA publication bias and small sample effect estimates. The funnel plots of included studies suggest a low likelihood of publication bias or small sample effects in this NMA. Appendix S10 lists the SUCRA probabilities for each intervention on vascular function outcomes, with higher SUCRA values indicating a higher probability of ranking the exercise intervention favorably in the network.

FMD results

Ten studies assessed FMD (N = 394). The results show that compared to the control group, CET (0.91 [0.44, 1.39], P < 0.01), INT (2.66 [1.41, 3.91], P < 0.01), RT (1.09 [0.19, 1.99], P < 0.05), and HYB (1.00 [0.41, 1.60], P < 0.01) all demonstrated significant improvement in FMD in postmenopausal women (refer to Table 2). Table 3 and Appendix S10 demonstrate that all interventions, including INT (SUCRA = 99.30), RT (SUCRA = 55.70), HYB (SUCRA = 51.00), and CET (SUCRA = 43.80), were more effective than the control group (SUCRA = 0.30) in improving FMD in postmenopausal women.

PWV results

Eighteen studies assessed PWV (N = 644). The results show that postmenopausal women who underwent CET (−0.48 [−0.85, −0.12], P < 0.05), INT (−0.79 [−1.26, −0.33], P < 0.01), RT (−0.41 [−0.77, −0.05], P < 0. 05), CT (−1.19 [−1.89, −0.49], P < 0.01), or HYB (−0.68 [−1.05, −0.31], P < 0.01) had a significant improvement in PWV compared to the control group (Table 2, N = 644). Table 3 and Appendix S10 demonstrate that all interventions, including CT (SUCRA = 93.00), INT (SUCRA = 71.50), HYB (SUCRA = 61.70), CET (SUCRA = 40.40), and RT (SUCRA = 32.90), were more effective than the control group (SUCRA = 0.40) in improving PWV in postmenopausal women.

AIx results

Five studies evaluated AIx (N = 163). The results show that compared to the control group, INT (−0.77 [−1.30, −0.24], P < 0.01) and HYB (−0.52 [−1.00, −0.03], P < 0.05) demonstrated significant improvement in AIx in postmenopausal women (refer to Table 2). Table 3 and Appendix S10 demonstrate that all interventions, including INT (SUCRA = 85.70), HYB (SUCRA = 75.10), CET (SUCRA = 48.80), and RT (SUCRA = 20.70), were more effective than the control group (SUCRA = 19.80) in improving AIx in postmenopausal women.

IMT results

Five studies evaluated IMT (N = 415). The network comparison did not show a significant reduction in IMT (refer to Table 2). The effectiveness ranking of different exercise modalities for improving IMT in postmenopausal women was as follows: CT (SUCRA = 81.90), CET (SUCRA = 68.70), HYB (SUCRA = 41.70). All intervention measures, except for RT (SUCRA = 21.80), were superior to the control group (SUCRA = 35.90), as shown in Table 3, Appendix S10.

NO results

Six studies evaluated the effect of NO in postmenopausal women (N = 177). The results indicate that CET (1.65 [0.67, 2.62], P < 0.01), INT (5.81 [4.42, 7.20], P < 0.01), and CT (4.21 [1.98, 6.44], P < 0.01) significantly improved NO compared to the control group (Table 2). The effectiveness ranking of different exercise modalities for improving NO in postmenopausal women was INT (SUCRA = 90.60), CT (SUCRA = 82.70), RT (SUCRA = 43.20), and CET (SUCRA = 29.10). All interventions outperformed the control group (SUCRA = 4.50), as shown in Table 3 and Appendix S10.

Strengths and limitations

The study has several strengths. Firstly, it utilized five types of exercise modalities that are easily accessible, widely applicable, and cost-effective. This provides motivation and convenience for postmenopausal women to engage in physical activity. Secondly, this study is the first to use NMA to assess the impact of various exercise modalities on five vascular function indicators in postmenopausal women. The study assesses the impact on vascular health by examining vascular structure, arterial endothelial function, arterial stiffness, and endothelial function. This analysis provides scientific evidence for exercise prescriptions and cardiovascular health management in postmenopausal women, helping to develop personalized and effective exercise regimens.

Limitations of this study include its focus on literature published only between 2000 and 2023 and its restriction to English-language publications. To enhance comprehensiveness, future research should expand the scope of literature inclusion.