The risk of preterm birth in women with uterine fibroids: A systematic review and meta-analysis.

Worldwide 15 million babies are born preterm each year [ 1 ]. Preterm birth is the most important cause of perinatal mortality and morbidity [ 2 ]. Approximately 65% of preterm births has a spontaneous onset and the remainder is medically indicated for maternal or fetal reasons [ 3 ]. The aetiology of preterm birth is multifactorial and complex [ 4 ]. Identifying risk factors for preterm birth is important to provide accurate counselling and it might create opportunities for improvement of antenatal care for women who are at risk. Previous reviews have suggested that women with fibroids have an increased risk of preterm birth [ 5 , 6 ]. Both associations with spontaneous and medically indicated preterm birth have been described. Several theories on the genesis of spontaneous preterm birth exist, such as a less distensible uterus and increased oxytocin levels, resulting in premature contractions and cervical change [ 7 , 8 ], Studies have indeed found women with fibroids have a shorter cervix [ 9 , 10 ]. Fibroids could also disturb uterine contraction patterns [ 11 – 14 ]. Furthermore, the hyperinflammatory state of fibroids could precipitate preterm labour. As described in women with endometriosis, this hyperinflammatory state could impair decidua/trophoblast interaction and uteroplacental development, which, in turn, could increase the risk of hypertensive disorders, placental abruption and fetal growth restriction, for which preterm birth might be medically indicated [ 15 ]. Studies have indeed shown an increased risk of these pregnancy complications in women with fibroids [ 5 , 6 , 16 – 19 ]. To quantify the risk of preterm birth in women with fibroids compared to women without fibroids, we performed a systematic review and meta-analysis. Furthermore, we aimed to differentiate spontaneous from medically indicated preterm birth, and to evaluate how fibroid characteristics (i.e. size, number, localisation) modify the risk of preterm birth.

The search provided 2078 unique articles which were assessed for eligibility ( Fig 1 ). One additional relevant article was found through cited-reference search, finally leading to the inclusion of 11 studies and 256,650 deliveries in this systematic review. Two publications reporting on an identical study population and were both used to complement the study data [ 25 , 26 ]. Characteristics of excluded studies which may appear to meet the inclusion criteria are available in S1 Table [ 27 – 38 ]. The included studies were conducted between 2009 and 2015. Five of the studies were cohort studies [ 9 , 16 , 18 , 19 , 25 , 26 , 39 ] and the other five were case-control studies based on the exposure of fibroids [ 10 , 40 – 43 ]. One study was prospective [ 43 ]. Study characteristics and sample sizes are demonstrated in Table 1 . A total of 256,683 singleton pregnancies with 256,650 deliveries were included in the systematic review: 12,309 in the fibroid group and 244,341 in the control group. The prevalence of fibroids within the cohort studies was 2.9% (6,185/221,813), ranging between 0.8% and 16.8%. All women received routine ultrasound screening ( Table 2 ). Two studies performed screening in the first trimester [ 16 , 39 ]. Egbe et al. included fibroids >3 cm and Arisoy et al. included intramural fibroids >3 cm [ 39 , 40 ]. Other studies did not specify a minimum measurement threshold in their exposure definition. Studies did not report whether they used 2D or 3D imaging with or without colour Doppler. a Prevalence of fibroids reported for non-fertility cohort studies only b This study reported a fibroid prevalence of 1.5% including operated fibroids, which were excluded in the present analysis. NS not specified Baseline characteristics of the included study populations are described in Table 3 . Women with fibroids were generally older and more often black, except in the studies that matched for age and ethnicity. In most studies, women with fibroids had lower parity. Three studies corrected for a previous preterm birth [ 16 , 18 , 19 ]. Blitz et al. excluded women with previous cervical surgery [ 9 ]. Eze et al. and Egbe et al. did not adjust for confounders by either statistical analysis or matching of characteristics [ 39 , 43 ]. BMI Body Mass Index; HDP hypertensive disease of pregnancy; GDM Gestational Diabetes Mellitus; NR not reported According to the Newcastle-Ottawa scale, seven of the included studies were of good quality [ 10 , 16 , 18 , 19 , 25 , 26 , 41 , 42 ], and four of poor quality [ 9 , 39 , 40 , 43 ]. See S1 Fig for the complete risk of bias assessment. The most important reasons for a fair or poor quality rating of studies were: not controlling or matching for important confounders such as maternal age and ethnicity, and high or unknown rates of loss to follow-up. The funnel plot of included studies has a fairly symmetrical pattern, indicating there is no clear evidence of publication bias ( S2 Fig ). Preterm birth was defined as birth <37 weeks of gestation by all studies and several studies made subgroups at other gestational age cut-offs. The lower limit of the gestational age cut-off mostly varied between 20–24 weeks of gestation. Zhao et al. and Egbe et al. used a cut-off at 28 weeks of gestation ( S2 Table ) [ 19 , 39 ]. Pooled preterm birth rates <37 weeks of gestation were 11.6% in the fibroid group and 9.0% in the control group ( Fig 2A ). The meta-analysis of 256,650 women showed women with fibroids had an increased risk of preterm birth <37 weeks compared to women without fibroids (OR 1.66, 95% CI 1.29–2.14). Heterogeneity was considerable (X 2 = 43.99, p<0.01; I 2 = 77%). Meta-analyses also showed a significant association of fibroids with preterm birth <34, <32, and <28 weeks of gestation and the strength of this association increased with preterm birth at earlier gestational ages ( Fig 2B–2D ). Meta-analyses for preterm birth <34 weeks of gestation included five studies with 109,457 women (5.2% versus 3.3%; OR 1.88, 95% CI 1.34–2.65), <32 weeks of gestation included three studies with 25,691 women (4.7% versus 2.6%; OR 2.03, 95% CI 1.40–2.95) and <28 weeks gestation included two studies with 25,418 women (3.0% versus 1.5%; OR 2.24, 95% CI 1.45–3.47). We did not identify any studies who reported on mid-trimester fetal losses between 16 and 22 gestational weeks. Meta-analysis of preterm birth <37 (A.), <34 (B.), <32 (C.) and <28 (D.) weeks of gestation in women with fibroids. * indicates studies that corrected for potential confounders. Only Girault et al. differentiated between spontaneous and medically indicated preterm birth <37 weeks of gestation. They found that women with fibroids are at increased risk of medically indicated preterm birth compared to women without fibroids (RR 1.99, 95% CI 1.18–3.36). The medical indications leading to these preterm birth were not reported. There was no difference in spontaneous preterm birth between the fibroid group (5/154 = 3.2%) compared to the control group (820/19,565 = 4.2%); RR 0.77 (95% CI 0.33–1.84) [ 16 ]. Conversely, Ciavattini et al. did show an increased risk of spontaneous preterm birth in women with fibroids (OR 2.33, 95% CI 1.11–4.88) [ 42 ]. Eight studies reported on PPROM, including 212,348 women: 6,122 in the fibroid group and 206,226 in the control group [ 10 , 16 , 18 , 19 , 25 , 26 , 41 , 42 ]. Meta-analysis showed no difference in PPROM rate between the fibroid and the control group (10.9% versus 9.9%; OR 1.87, 95% CI 0.96–3.64) ( Fig 3 ). Statistical heterogeneity was substantial (X 2 = 14.96, p = 0.04; I 2 = 53%). An overview of the crude and adjusted effect sizes per study and outcome can be found in S3 Table . * indicates studies that corrected for potential confounders. Good quality studies were included in sensitivity analyses ( S3 Fig ). All these studies adjusted for potential confounders. The meta-analyses showed a smaller, yet still significant, association between fibroids and preterm birth <37 (11.5% versus 9.1%; OR 1.50, 95% CI 1.15–1.95) and <34 (4.7% versus 3.2%; OR 1.50, 95% CI 1.16–1.94) weeks of gestation. A meta-analysis of preterm birth <32 and <28 weeks of gestation could not be performed as there was one good quality study assessing these outcomes. There was no significant association between fibroids and PPROM in a sensitivity analysis (11.3% versus 10.5%; OR 1.97, 95% CI 0.84–4.61). Girault et al. was rated as good quality but was not included in sensitivity analyses for preterm birth <34 weeks of gestation and PPROM [ 16 ]. They provided additional preterm birth rates in their unoperated fibroid group upon our request. As we did not have the adjusted ORs for these particular outcomes, this lead to the deduction of points in quality in the comparability section with regards to these outcomes. The results of the subgroup analyses based on fibroid characteristics are shown in S4 Fig . Subgroup analysis including studies comparing women with fibroids >3 cm and >5 cm to women without fibroids showed that, overall, there was a significant association between fibroids and preterm birth <37 weeks of gestation (n = 177,521; OR 2.03, 95% CI 1.26–3.27). Heterogeneity was substantial (X 2 = 13.29, p = 0.02; I 2 = 62%). Meta-analyses did not show differences in preterm birth 5 cm) to small (≤5 cm) fibroids (OR 1.38, 95% CI 0.78–2.46), nor when comparing multiple to single fibroids (OR 1.89, 95% CI 0.79–4.53). No subgroup analyses could be performed on fibroid location (submucosal, intramural, subserosal), vascularisation, FIGO classification or other ultrasound features. Neither could we perform subgroup analyses based on parity, a history of previous mid-trimester fetal losses or preterm birth, or a history of dilation and curettage.

This systematic review and meta-analysis suggests fibroids are associated with total preterm birth and this association is stronger with earlier gestational age of the preterm birth. However, the considerable amount of heterogeneity between studies may indicate biased results. Considering the magnitude of the disease burden of preterm birth, as well as the biological plausibility of an association between preterm birth and fibroids, we encourage further research to clarify this association through prospective and systematic myometrial assessment in early pregnancy. Finally, it is important that these studies distinguish between spontaneous and medically indicated preterm birth.

This systematic review was reported according to the PRISMA guidelines for systematics reviews and meta-analyses ( S1 Appendix ) [ 20 ]. No funding was received for the conduct of this systematic review and meta-analysis. Ethical approval was not required for secondary use of data in this systematic review and meta-analysis. The protocol was registered in the PROSPERO database in July 2020 (CRD4202018697]. After registration, the following changes were made to the protocol: 1) we only included studies that performed routine ultrasound/imaging for the diagnosis of fibroids to reduce the risk of ascertainment bias; 2) we removed additional outcomes such as preeclampsia as these were not incorporated in the search terms; and 3) we changed the statistical analysis technique from a random-effects model to the generic inverse-variance method to include confounder-adjusted estimates. A systematic literature search was performed by a medical information specialist (JK) and two researchers (GV and AL) on 9 June 2021 in the following databases: PubMed, Embase and Web of Science. The search terms included both keywords as well as free text terms for ‘fibroids’ and ‘preterm birth’, along with their synonyms. The search strategies are presented in S2 Appendix . No language, publication date or other restrictions were applied. All duplicates were excluded using EndNote. A cited-reference search was performed to identify potential additional relevant studies. Original studies reporting on the risk of preterm birth in women with and without fibroids were eligible for the present systematic review and meta-analysis. To limit the risk of the opportunistic detection (i.e. incidental identification) of fibroids, the fibroids had to be diagnosed by routine ultrasound or other imaging before or during pregnancy, indicating we excluded studies that, for example, included women that were admitted to the hospital or selectively referred for ultrasound screening. We also excluded studies including women with multiple gestations due to their increased baseline risk of preterm birth, and studies reporting on preterm birth after surgical or non-surgical treatment of fibroids as this treatment may have influenced obstetric outcome. Furthermore, conference abstracts, case reports, case series, reviews and guidelines were excluded. For the purpose of feasibility, we excluded non-English articles. The main outcomes were preterm birth <37, <34, <32 and <28 weeks of gestation. In addition, we aimed to differentiate between the spontaneous onset of preterm birth following either the onset of contractions or rupture of membranes, and medically indicated preterm birth due to maternal or fetal complications. Other relevant outcomes were preterm prelabour rupture of membranes (PPROM) and mid-trimester fetal loss (between 16 and 22 weeks of gestation). Two authors (GV and AL) independently performed screening of all potential articles using Rayyan [ 21 ]. First screening was based on title and abstract. In case of inconsistency between authors, the full article was screened. We screened the full text of the remaining articles based on the aforementioned predefined in- and exclusion criteria for eligibility. Disagreements between the authors were resolved by discussion or consultation of a third author (MdB). We used the Newcastle-Ottawa Quality Assessment Scale (NOS) for nonrandomised studies to evaluate the validity of the included studies [ 22 ]. We slightly adapted the questions and answers to obtain a better fit for our studies. Each item could be given a maximum of one star. Studies were categorised as good, fair and poor quality according to the Agency for Health Research and Quality standards. Our customised NOS scale and thresholds for quality assessment are demonstrated in S3 Appendix . The final judgement of quality was summarised in a risk of bias table. No assessment of certainty was planned in the protocol. Two authors (GV and AL) independently extracted data from the included articles and performed quality assessment. We collected data on author, year and country of publication, language, publication status, start date and end date of the study, ethical approval, type and source of funding, study design, sample size, population characteristics (e.g. age, ethnicity, parity, obstetrics history, previous cervical cerclage surgery), diagnostics method used for the identification of uterine fibroids, fibroid characteristics (size, location, number, vascularisation, FIGO classification), past treatment of fibroids (surgical or non-surgical), the outcome measures as described previously, and key conclusions. Corresponding authors of the included articles were contacted for further information if relevant outcomes or other data were not presented in the original publication. Discrepancies in data extraction and quality assessment were resolved by discussion and consulting a third author (MdB). Outcome event rates and crude odds ratios were calculated based on the number of deliveries. If studies included miscarriages in their results, these were excluded from the sample size. If studies were unclear about in- or exclusion of miscarriages in their sample size, we used the reported sample size for our calculations. The meta-analysis included adjusted odds ratios (ORs) of those studies that corrected for confounders and crude ORs of studies that did not. We used a random-effects model with the generic inverse-variance method [ 23 , 24 ]. As we included both cohort and case-control studies, the results of the meta-analyses were presented as odds ratios and 95% confidence intervals (95% CI). Heterogeneity between studies was evaluated by using X 2 and I 2 statistics and possible publication bias was assessed with a funnel plot. We performed a pre-planned sensitivity analysis of good-quality studies. In addition, we planned to perform subgroup analyses based on the size, number, location and vascularisation of the fibroids; parity, a history of previous miscarriages or preterm birth, and a history of dilation or curettage. Statistical analyses were performed using Review Manager (RevMan 5.3.5) and R Studio (version 4.0.3: meta, metafor and dmetar packages).

(PDF) Click here for additional data file. (PDF) Click here for additional data file. (PDF) Click here for additional data file. (PDF) Click here for additional data file. (PDF) Click here for additional data file. (PDF) Click here for additional data file. (TIF) Click here for additional data file. (TIF) Click here for additional data file. Sensitivity analyses of good quality studies for preterm birth (A) <37 and (B) 3 and >5 cm compared to women without fibroids; (B) large versus small fibroids; and (C) multiple versus single fibroids. * indicates studies that corrected for potential confounders. (TIF) Click here for additional data file.