Uterine volume assay after gonadotoxic therapies in childhood, adolescence, and young adulthood: A systematic review and Bayesian network meta-analysis.

Eloïse Fraison: designed the work, selected the studies, extracted the data, wrote the manuscript. Stephanie Huberlant: designed the work, selected the studies, extracted the data. Mathilde Cavalieri, Aurore Gueniffey, Justine Riss, and Christine Rousset‐Jablonski: extracted the data. Blandine Courbiere: designed the work, corrected the manuscript. All authors analyzed the results and approved the final version of the manuscript.

The study was funded by an unrestricted grant from GEDEON RICHTER, France.

Our search strategy revealed 2847 reports, of which 519 were duplicates. After screening the titles and abstracts, 158 reports were potentially eligible and retrieved in full text (Figure  1 ). Four articles were included in this meta‐analysis. 2 , 3 , 4 , 16 Uterine volume was assessed using ultrasonography in three articles and MRI in one article. All studies were published in English, and their main characteristics are summarized in Table  1 (Table  S1 for excluded studies). PRISMA flow diagram of study selection. Main characteristics of studies included in the systematic review and network meta‐analysis. Inclusion criteria : female survivors diagnosed and treated for a noncentral nervous system childhood cancer (malignant hematologic disease, solid tumor); ≤15 years old at the time of diagnosis, treated with radiotherapy and/or chemotherapy; off treatment for at least 1 year at study inclusion; in complete remission, ≥ 18 years old at study inclusion. Exclusion criteria : lost to follow‐up, previous hysterectomy, institutionalized mental disorders Inclusion criteria : women survivors of childhood malignancies and patients who have received SCT for thalassemia major or sickle cell anaemia. Exclusion criteria : N/A Treatment : 1984–2011 Assessment : January 2010–December 2012 Inclusion criteria : women >18 years old participating in the LEA cohort; those who underwent HSCT during childhood or adolescence for acute leukemia and after an MAC regimen including either TBI or an alkylating agent‐based regimen; who agreed with undergoing pelvic magnetic resonance imaging (MRI) to assess their uterine volume. Exclusion criteria : N/A Inclusion criteria : women enrolled in the radiology department of the principal investigator for pelvic MRI during the study period, aged between 18 and 40 years, without any uterine pathology who underwent pelvic MRI during the study period. Exclusion criteria : controls were excluded in case of uterine pathology, such as adenomyosis or fibroids Inclusion criteria : adult women treated for a cancer before the age of 18 years who survived for at least five years after diagnosis and were at least 18 years old at study entry with two inclusion criteria: they had a transvaginal ultrasound scan performed and a blood sample taken their former childhood cancer treatment included external beam RT to a field that certainly (total‐body irradiation and pelvic RT) or most likely (low‐abdominal and lower spinal RT) involved part of the uterus they had a transvaginal ultrasound scan performed and a blood sample taken their former childhood cancer treatment included external beam RT to a field that certainly (total‐body irradiation and pelvic RT) or most likely (low‐abdominal and lower spinal RT) involved part of the uterus Exclusion criteria : Childhood cancer survivors without a uterus or with uterine fibroids Childhood cancer survivors without ultrasonographic data not stored Childhood cancer survivors without a uterus or with uterine fibroids Childhood cancer survivors without ultrasonographic data not stored Treatment : 1963 and 2002 Assessment : January 2008–July 2014 Most studies were classified as having a low‐to‐moderate risk of bias. Two studies had a moderate risk of participant selection bias owing to a lack of detailed information about the patient selection process. 2 , 3 For intervention classification, two studies had a moderate risk of bias because the chemotherapy protocols lacked sufficient detail. 2 , 16 Regarding deviation from the intervention, one article was classified as having a moderate risk of bias because the description of the intervention was not clear. 2 For missing data, one study was classified as having a moderate risk of bias because the number of patients differed between the table and text. Uterine volume data were available for 225 women after chemotherapy, 153 women after chemoradiotherapy, and 257 women without cancer (controls). The results with random and fixed effects were similar; therefore, only random results were presented. Uterine volume was significantly lower in the chemoradiotherapy group than in the control group (−29.2 mL [−49.1 to −12.5]) and compared to the chemotherapy group (−20.9 mL [−39.1 to −0.3]). The MD in uterine volume between the control and chemotherapy groups was 8.2 mL [−11.8–34.2] and not significant (Figure  2 ). The results were similar after the exclusion of outliers—e.g., the control group in the study by Van de Loo et al. and the parous chemotherapy group in the study by Courbiere et al. reinforcing the same differences and trends (Figures  3 and 4 ). Uterine volume: Bayesian NMA Forest plot, with parity as a dummy variable (mln). Uterine volume: Bayesian NMA network graph after the exclusion of outliers (mln). Uterine volume: Bayesian NMA Forest plot, with parity as a dummy variable without outlier (mln). SUCRA for the control, chemotherapy, and chemoradiotherapy groups were 0.91, 0.57, and 0.01, respectively (Table  2 ). One study did not display data by parity ( n  = 165), had a serious risk of bias, and was excluded from the sensitivity analysis. SUCRA showed the same ranking, and sensitivity analyses led to the same conclusions. Uterine volume, SUCRA, Bayesian (mln). Note : SUCRA: surface under the cumulative ranking distribution defined in Salanti et al. (2011). Order according to surface, with first rank corresponding to the highest surface. Only three studies reported results according to parity, including 387 nulliparous and 103 parous women. In nulliparous women, NMA showed moderate heterogeneity or inconsistency (τ 2  = 72.6, I 2  = 51.6% [0%; 84%]). There was no significant difference in uterine volume between the chemotherapy and control groups (−2.4 mL; [−19.1–14.3]). The uterine volume in the chemoradiotherapy group was significantly lower than in the control group (−21.6 mL; [−39.5 to −3.7]) and the chemotherapy group (−19.2 mL; [−31.6 to −6.7]) (Figures  5 and 6 , Table  3 ). Uterine volume, Nulliparous: Forest plot of direct and indirect evidence and network (random effect, ref. = control group, mln estimates, frequentist NMA). Uterine volume, Nulliparous: Forest plot of direct and indirect evidence and network (random effect, ref. = chemotherapy only, mln estimates, frequentist NMA). Uterine volume in nulliparous: League table (mln estimates, fréquentiste NA, random‐effects model). Note : Data are difference in Means (MD) of UV with 2‐sided 95% confidence interval. Pairwise comparisons of the therapy in the column versus the therapy in the row in the lower triangle and row versus column in the upper triangle. Estimates from network meta‐analysis in the lower triangle and the direct treatment estimates from pairwise comparisons in the upper triangle. SUCRA for the control, chemotherapy, and chemoradiotherapy groups were 0.8, 0.7, and 0.006, respectively (Table  4 ). Uterine volume, SUCRA value (mln estimates, frequentist NMA). In parous women, the NMA was performed and showed high heterogeneity or inconsistency (τ 2  = 496.1, I 2  = 55.4% [0%; 85.2%]). There was no significant difference in uterine volume between the chemotherapy and control groups (−16.9 mL; [−57.5–23.6]) and the chemoradiotherapy and control groups (−31.4 mL; [−71.9–9]). Moreover, in parous women, there was no significant difference between the chemoradiotherapy and the chemotherapy groups in uterine volume (−14.5 mL; [−47.3–18.3]) (Figures  7 and 8 , Table  5 ). Uterine volume, Parous: Forest plot of direct and indirect evidence and network (random effect, ref. = control group, MLN estimates, frequentist NMA). Uterine volume, Parous: Forest plot of direct and indirect evidence and network (random effect, ref. = chemotherapy only, MLN estimates, frequentist NMA). Uterine volume in parous: League table (mln estimates, fréquentiste NA, random‐effects model). Note : Data are difference in means (MD) of UV with 2‐sided 95% confidence interval. Pairwise comparisons of the therapy in the column versus the therapy in the row in the lower triangle and row versus column in the upper triangle. Estimates from network meta‐analysis in the lower triangle and the direct treatment estimates from pairwise comparisons in the upper triangle. SUCRA for the control, chemotherapy, and chemoradiotherapy groups were 0.81, 0.68, and 0.005, respectively (Table  4 ).

In this meta‐analysis, a significantly lower uterine volume was observed in the chemoradiotherapy group compared to the control group. However, no significant difference in uterine volume was observed between the chemotherapy and control groups, even after adjustment for parity. The negative impact of chemoradiotherapy on the uterus is already well documented 1 and our results align with those reported in cohort studies of cancer survivors. 2 , 3 , 4 , 16 However, recent studies have raised a negative impact of high‐dose chemotherapy alone with a significant reduction of uterine volume. 2 , 3 , 4 For instance, Courbiere et al. observed a significant decrease in uterine volume in a homogeneous cohort of acute leukemia female survivors exposed during childhood and adolescence to standardized conditioning regimens with high doses of alkylating agents. 4 However, it remained difficult to determine whether the significant reduction in uterine volume observed with high‐dose alkylating chemotherapy was due to the direct cytotoxicity of the alkylating agents themselves or whether it was an indirect consequence of chemotherapy‐induced ovarian damage and subsequent hypoestrogenism. Van de Loo et al.'s study also reported that the uterine volume of nulliparous women exposed to chemotherapy was lower compared to the control groups, with a median of 48.1 mL (35.7–61.8 mL) versus 61.3 mL (49.1–75.5 mL). Nevertheless, after having corrected the analyses using the alkylating agent dose score, they did not observe a significant risk of having a small uterine volume in chemotherapy‐exposed nulliparous childhood cancer survivors (1.34, 95% CI 0.52–3.70). Parous women also did not show differences between the exposed groups in the risk of having a small uterine volume (<63.1 mL). Notwithstanding, uterine volumes were 68.1 mL (47.3–84.9 mL) and 90.2 mL (66.6–106.4 mL) in the post‐chemotherapy and control groups, respectively, and the threshold of 63.1 mL for defining a small uterus was an arbitrary threshold. These results led us to perform our metanalysis, and we observed that when the data were pooled, no significant differences were found between the control and chemotherapy groups regarding uterine volume. The major strength of our study was its rigorous methodology, with a strict selection of included studies. We only included studies with a control group because the estimation of uterine volume can vary depending on the operator, imaging technique, and mathematical equations used. Therefore, volume data from cohort studies without a control group are likely to introduce excessive bias. The risk of assessment bias was determined by using the Cochrane ROBINS‐I tool. Sensitivity analyses were performed when required. The study population was homogeneous in terms of age because we selected studies that focused on AYA. However, this study has some limitations. First, the measurement of uterine volume was not homogeneous, as it was assessed through ultrasonography in three articles 2 , 3 , 16 and through MRI in one. 4 Only four articles could be included, which was sufficient to perform a meta‐analysis. However, with this heterogeneity of measurement and the low number of included articles, the results should be interpreted with caution. Moreover, we had to adapt the results to those of other studies. Only Courbiere et al. presented the results using means, whereas the other three studies showed medians. Therefore, we had to convert the medians into means to be able to perform a meta‐analysis. In some of the included articles, we had to divide the effectiveness to ascertain the gonadotoxic treatment received in each group of patients. This division reduced the number of included patients in each group and, therefore, the strength of our study. Finally, regarding the comparison of the chemotherapy and control groups, only three publications had a specific group treated with chemotherapy alone that could be compared to healthy women, and the article by Van de Loo et al. held more weight than the others. 2 , 3 , 4 In this study, female patients were treated for a primary malignancy before 18 years of age, and the included population was heterogeneous, with numerous types of cancer and chemotherapy regimens. Among them, 78% of the patients had been exposed to chemotherapy before menarche, and only 61% were exposed to alkylating agents. This heterogeneity of the main study included in the meta‐analysis also led us to carefully interpret the results. Finally, when we look at the results according to parity, the small sample size included in the subgroup analysis again did not allow for any definitive conclusion regarding the influence of chemotherapy on uterine volume.

Our meta‐analysis confirmed the well‐known data on the damage caused by chemoradiotherapy to the uterus but did not find any significant impact of chemotherapy alone on uterine volume. Our results contradicted some recent studies that reported a reduction in uterine volume after high‐dose chemotherapy. Sunguc et al. reported an increased risk of prematurity in a cohort of female survivors of adolescent and young adult cancer for all chemotherapy types. 17 Based on this study and the observations by Meirow et al. on the effect of cyclophosphamide on ovarian stroma, we hypothesize that chemotherapy could also damage the uterus despite a maintained uterine volume. 5 Moreover, studies in mice have shown damage to the skeletal muscle with muscle mass atrophy after cisplatin treatment. 18 Chemotherapy‐induced damage to the myometrium has never been studied in humans, and its effect on uterine muscle has never been demonstrated. It could be interesting to follow pediatric cohorts with homogenous groups of female patients (e.g., type of cancer, type of chemotherapy, doses of chemotherapy, hormonal follow‐up, and treatment) to provide insights into uterine and obstetrical impacts. Gathering the data between countries to provide specific answers for each oncological protocol with good statistical power could be valuable.

Uterine damage has already been well described after pelvic radiotherapy or total‐body irradiation, independently of ovarian effects. The consequences are decreased uterine volume, damage to the endometrial architecture, and impairment of uterine vascularization, leading to higher obstetrical morbidity. 1 Recent studies have suggested that chemotherapy alone in adolescents and young adults (AYA) may also be responsible for uterine damage. Indeed, Van de Loo et al. observed that uterine volume was significantly smaller in a group of childhood cancer survivor women exposed to chemotherapy compared to healthy women (odds ratio [OR] 2.61, 95% confidence interval [CI], 1.15–5.91). 2 In a population of women treated during childhood with hematopoietic stem cells transplantation (HSCT), Beneventi et al. reported that uterine volume was significantly reduced by 81.9% (95% CI, 71.8–87.8) after total‐body irradiation and by 67.4% (95% CI, 58.5–75.6) after a conditioning regimen with a high dose of alkylating agents. 3 Courbiere et al. observed similar results after HSCT with a significant uterine volume reduction of 43.1% (28.8–57.4) after alkylating agent‐based conditioning compared to a control group of healthy women. 4 Many hypotheses could explain the physiopathology of reduced uterine volume after chemotherapy; however, none have been confirmed as of yet. These include fibrotic damage induced by high‐dose chemotherapy, 5 the impact of hormonal deprivation around puberty, and a lack of compliance with hormonal replacement therapy. 6 Importantly, the reduced uterine volume could be responsible for adverse pregnancy outcomes, such as spontaneous abortion, preterm birth, low birth weight, or small for gestation age birth. 1 The objective of this systematic review and meta‐analysis was to assess the long‐term impact of exposure to gonadotoxic therapies during childhood, adolescence, and young adulthood (CAYA) on uterine volume.

The authors declare no conflicts of interest.

The study protocol was prospectively registered in PROSPERO (CRD42023440219), and the reporting of this systematic review and meta‐analysis was guided by the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses for Network Meta‐Analysis (PRISMA‐NMA) guidelines. We only included comparative studies published as full‐text papers involving female cancer survivors or women who underwent HSCT, were aged over 18 years, and had received chemotherapy and/or radiotherapy during childhood (<15 years), adolescence, or young adulthood (15–25 years). Studies were included if they reported uterine volume and compared two or more of the following groups: chemotherapy alone, radiotherapy, chemoradiotherapy, and controls. Radiotherapy was defined as total‐body or pelvic irradiation. We excluded animal studies, as well as studies involving women treated for cancer or receiving HSCT after the age of 25 years, displaying results only for age, or with a median age exceeding 20 years. Additionally, studies that included radical trachelectomy only for cervical cancer, anticancer treatments involving hysterectomy, or chemotherapy for gestational trophoblastic neoplasia were excluded. Uterine volume was defined as the volume of the uterus assessed using abdominal or pelvic ultrasonography or magnetic resonance imaging (MRI). The Cochrane Library, Embase, and PubMed databases were searched from inception until April 30, 2023. The search was limited to articles published in English or French. We used a combination of Medical Subject Heading terms and text words, including “cancer survivors,” “bone marrow transplantation,” “chemotherapy,” “radiotherapy,” and “uterine volume.” The full search strategy for each database is described in detail in the appendix. Relevant reviews identified in the search were screened to identify additional relevant citations (Supporting Information Appendix  S1 ). Two independent reviewers conducted the initial screening of the titles and abstracts of all articles without information on the authors, institutions, journal titles, and study results. The full texts of the potentially relevant articles were independently retrieved and assessed for inclusion by two reviewers. The reasons for the exclusion of ineligible studies were documented. Methodological validity was also assessed before inclusion in the review using the adapted Newcastle–Ottawa Quality Assessment Scale for cohort and case–control studies. 7 Any discrepancies or uncertainties were resolved through discussions among the reviewers to reach a consensus. Pairs of reviewers independently extracted data from the included articles using a data‐extraction form designed by the authors. The following details were collected for all included studies: country, study design, study type, inclusion and exclusion criteria, study period, cancer and therapy characteristics, age at main therapy, follow‐up duration, age at examination, parity, comorbidities, and outcomes. For articles published after 2013 that included women under the age of 25 years, the authors of the original studies were contacted to collect any pertinent missing data if the primary outcome was reported overall but not specifically for each treatment group. Two independent reviewers assessed the methodological quality of the selected studies. The risk of bias was assessed using the ROBINS‐I tool. 8 Box–Cox, quantile estimation, and method for unknown non‐normal distribution (MLN) approaches were used to estimate the sample mean and standard deviation (SD) when not documented in an article. 9 , 10 No other replacements for missing data were performed. A Bayesian NMA was performed to quantify the relative association of each therapy with uterine volume adjusted for parity, assuming similarity and transitivity. 11 Articles with uterine volume documented by parity were split into nulliparous and parous and were considered independent articles in the NMA. However, if the results for a given therapy were reported in several groups (other than parity), then only the group with the highest number of subjects was included in the NMA (e.g., Ref. [ 3 ]). Convergence was assessed using Gelman–Rubin diagnostic statistics, plots, and autocorrelation plots. 12 After generating the network geometry, local inconsistency was estimated by comparing the direct and indirect evidence using the node‐splitting method. Heterogeneity and global inconsistencies were assessed using the Q ‐test and I 2 statistic. 11 To better control for heterogeneity across studies, random‐effect models were used under the assumption of consistency. The relative effect size was presented as the difference in means (MD) with 95% credible intervals. Model fit was assessed using the overall residual deviance (Dbar) and leverage plots. Dbar plots were used to detect outliers. Therapies were ranked using the surface under the cumulative ranking curve (SUCRA). 13 Funnel plots could not be used to evaluate publication bias because only a few studies were included in the NMA. In the second step, because the intensity of the relative association of each therapy with uterine volume possibly varied with parity, a frequentist NMA was performed independently in parous and nulliparous women. The relative effect size was presented as MD with 95% CI. Heterogeneity and global inconsistencies were assessed using the Q statistic, tau 2 , and I 2 . Direct and indirect evidence were split using a back‐calculation method and compared. Sensitivity analyses were performed using the mean and SD estimates of each approach, fixed‐effects models, exclusion of possible outliers, and exclusion of articles with a serious risk of bias. Statistical analyses were performed using R (version 4.3.2), the R package estmeansd (version 1.0.1) to derive estimations of means and SD not documented in articles, bnma (version 1.6.0) using the Markov Chain Monte Carlo engine JAGS (version 4.2) for Bayesian NMA, and netmeta (version 2.9.0) for frequentist NMA. 9 , 14 , 15 The quality of evidence for each outcome was determined through the grading of recommendations, assessment, development, and evaluation working group methodology.

Appendix S1. Table S1.