Non-malignant gynaecological disease and risk of cardiovascular or cerebrovascular disease: a systematic review and meta-analysis

Discussion

This systematic review and meta-analysis suggests an association between NMGD and C/CVD. The meta-analysis comprises a large total sample size allowing exploration of several subtypes of NMGD and C/CVD; the 28 studies yield a total of 3 271 242 individuals. The primary outcome indicates an increased risk of developing composite C/CVD in individuals with at least one NMGD (the exposed population), in comparison to individuals without NMGD (the unexposed population), although high study heterogeneity and risk of bias prevalence were identified. Sub-group analyses correspondingly suggested that the increased risk of C/CVD and its components was observed in individuals with endometriosis or PCOS. Given the importance of temporality for causal inference, we further conducted sensitivity analyses that showed no significant difference in effect estimates when considering only studies that took into account temporality, although the number of studies contributing to this comparative analysis was small. In addition, when excluding the most influential studies, we found no substantial influence of any of the included studies.

Common biological pathways linking NMGD to C/CVD may include systemic inflammation and endogenous estrogenic milieu. The categorisation of atherosclerosis as a chronic low-grade inflammatory condition is now commonly accepted.45 Inflammation and oxidative stress are important triggers for C/CVD, and various inflammatory markers have been found to be essential contributors to C/CVD.46 47 When considering endometriosis specifically, the mechanism believed to link this condition to increased cardiovascular risk is systemic inflammation.48 Similarly, in PCOS, a cross-sectional analysis found a significant association between low-grade inflammation and sympathetic dysfunction and hyperandrogenism.49 As such, inflammation may be an underlying mechanism for an association between NMGD and C/CVD. However, there is also a possible overlap of gynaecological risk factors and cardiovascular risk factors that may act as confounders. For instance, features of metabolic syndrome, a complex interaction of visceral obesity, dyslipidaemia, hyperglycaemia and hypertension50 has been consistently shown to be present in individuals with PCOS.51 Metabolic syndrome is associated with an increased risk of C/CVD events and mortality.52

Female steroid hormones are implicated in the development of NMGD and link these to cardiovascular risk factors.48 Endogenous estrone is related to endothelial function, while estradiol is related to vascular remodelling, suggesting specific roles for different female steroid hormones.53 The effects on endothelial function seem beneficial, providing cardiovascular protection to individuals AFAB.54 While some studies adjusted their analysis for menopausal hormone therapy (MHT) use18 22 27 or oral contraceptive use,23 27 32 34 only one35 performed a sub-group analysis, which demonstrated that the risk of subsequent CAD was higher in patients with endometriosis taking hormonal therapy. Similarly, hysterectomy is associated with a greater risk of C/CVD,55 which could be further exacerbated when bilateral oophorectomy is performed.56 Although few studies have considered this potential mediation,23 57 those that have assessed this have found a significant proportion of the causal pathway associating NMGD with C/CVD to be explained by surgical treatments. However, independent variation remained, confirming that other pathways contribute to C/CVD risk among individuals with NMGD.

The results of this review must be interpreted with caution given the limitations. First, most of the findings in the meta-analysis suggest moderate to very high heterogeneity among studies, which may yield imprecise meta-analytic summary estimates. Indeed, there were a few examples of heterogeneous results when comparing single studies. Saavalainen et al
25 found that the risk of C/CVD mortality was lower in endometriosis patients compared with controls; the underlying cause of this discrepancy is unclear. Furthermore, Solomon et al
27 reported a lower risk of IHD in the irregular menstrual cycle group compared with those reporting regular cycles. The participants were asked to recall and self-reported menstrual cycle irregularity at ages 20–35. This methodology may contribute to inaccurate categorisation, although validation studies conducted within this cohort suggested that the self-reporting of medical conditions was reliable.58

The second limitation is the risk of bias in many of the studies. Concerningly, 53.5% of the studies included in this meta-analysis had low design quality with ‘serious’ or ‘critical’ risk of bias. The primary driver of poor risk of bias scores was a lack of control for potential confounders. While observational studies require data adequate to quantify and statistical approaches that account for potential confounders to support valid causal inference,59 we found that a large proportion of the studies published to date did not include any methods for control or assessment of confounding factors at all. It is important to note, however, that the higher quality studies with a low risk of bias (just six studies) that did account for potential confounders observed little evidence of confounding with low magnitudes of confounding effects.23 25 34 35 39 41 The ROBINS-I risk of bias tool11 was designed for non-randomised interventional studies; therefore, applying it in this systematic review was potentially less precise.

Furthermore, despite lenient inclusion criteria, the data were sparse for most outcomes. To conduct a robust meta-analysis, outcomes were combined from a clinical perspective, reflecting how these diagnoses are managed in clinical practice. The composite C/CVD outcome reported by the studies was defined within the study, which meant that our primary outcome could not be broken down by C/CVD subtype, although most studies reported a separate outcome for each C/CVD subtype which was reported in our subgroup analyses. No studies reporting the association between atrial fibrillation and NMGD were eligible for inclusion in this review. There was a paucity of studies reporting effect estimates for the association between peripheral vascular disease and NMGD, which meant that this outcome could not be meta-analysed. Including these composite C/CVD subtypes in the meta-analysis would allow for more robust results regarding the association examined. Similarly, the only NMGD subtypes examined by the included studies were endometriosis, PCOS, dysmenorrhea and irregular menstrual cycles.

Additionally, five studies utilised data from the Taiwan National Health Insurance Database,33–35 38 43 which could potentially inflate the total number of individuals included in this review if an individual appeared in the dataset of more than one study. This limitation was addressed by conducting a subgroup analysis including only one Taiwan National Health Insurance Database study per condition, which was not significantly different from the main outcome. A further limitation of this review was that a couple of studies split the exposure group into subgroups based on menstrual regularity27 or PCOS phenotype,60 which may contribute to misclassification bias. In contrast, multiple studies17–20 22 28 30 32 33 36 37 39–42 did not differentiate between ischaemic and haemorrhagic stroke subtypes, which have distinct underlying pathophysiological mechanisms.61 When considering the population included in this study, it is possible that individuals with NMGD were more likely to seek medical attention or to be referred for C/CVD investigations as they had a known health condition than control groups. This may have led to a greater diagnosis of C/CVD in the exposure group and is a limitation to consider when interpreting the results of this review.

Lastly, we pooled different risk estimates, which may overestimate or underestimate the observed association between NMGD and outcomes; therefore, it is important to consider these findings with caution. Ideally, only prospective studies that reported ORs or hazard ratios would be pooled in the meta-analysis; however, as there are few studies available on this topic, our review aims to provide a global view of the association. Despite these limitations, this systematic review and meta-analysis represents the highest level of evidence currently available, with a rigorous methodology and a large patient population for many of the associations explored, and a thorough meta-analysis with multiple subgroups and sensitivity analyses.

The results of this systematic review and meta-analysis may have implications for clinical practice. Although the extent of this association is still to be explored, and causality has not been established, the findings suggest that it is important to raise awareness of the potential association between NMGD and C/CVD both in the general public and healthcare professionals. Awareness of this association would allow healthcare professionals to advise patients regarding risk-reducing behavioural changes and interventions, to potentially prevent or delay the onset of or reduce the severity of C/CVD. Furthermore, healthcare professionals would be aware to monitor their patients with NMGD for early signs and symptoms of C/CVD, and potentially diagnose atypical presentations of C/CVD with the knowledge of the underlying risk for this group of patients. Management strategies for C/CVD in this patient group could be tailored to the hypothesised underlying mechanisms that connect these conditions: systemic inflammation and female steroid hormones.

The association between C/CVD and NMGD requires further exploration with high-quality longitudinal studies adjusted for confounders to establish temporal relationships and causality. Future research should focus on incidence or mortality rates as more studies are published on this topic. It is essential to establish an understanding of the mechanistic pathways linking the conditions, which would allow investigation of the potential modifiers of risk as well as targeted management plans. A larger body of literature with larger sample sizes within each study will afford much more precise estimation and the ability to explore likely variation among all NMGD and C/CVD subtypes. Additionally, it would be beneficial to standardise the reporting of both NMGD conditions and C/CVD outcomes.

In conclusion, our findings suggest an association between NMGD and C/CVD overall and in subgroup populations. Given the limitations of the review, high study heterogeneity and risk of bias prevalence, future research must require study design rigour and improved NMGD and C/CVD harmonisation, and elucidate risk variation among subpopulations. Physicians should be aware of the potential association between NMGD and C/CVD and use this to inform clinical practice in order to mitigate the risk of C/CVD. Our findings reinforce the need for prospective longitudinal lifecourse research evaluating risk of C/CVD outcomes in NMGD overall and in subgroup populations, which may catalyse primary prevention strategies.