Fertility‐sparing treatment for atypical endometrial hyperplasia and endometrial cancer

Abstract

Background

Endometrial cancer is the sixth most common cancer in women worldwide, and the fourth most common in high‐income countries, where its incidence is increasing. Atypical endometrial hyperplasia (AEH) is an overgrowth of the womb lining and can be a precursor of endometrial cancer. Between 14% and 25% of cases of endometrial cancer are diagnosed in premenopausal women. Due to delays in childbearing age and increasing obesity rates, a growing number of women wish to explore fertility‐sparing management of endometrial cancer or AEH.

Objectives

To compare the effectiveness and safety of fertility‐sparing treatments, including pharmacological interventions (e.g. oral progestin, levonorgestrel intrauterine system (IUS), metformin) and bariatric or hysteroscopic surgery, for AEH and presumed stage IA grade 1 endometrioid endometrial cancer.

Search methods

We searched the following electronic databases to 3 February 2025: CENTRAL; Ovid MEDLINE; and Ovid Embase. We also searched five trials registers and conference proceedings and abstracts.

Selection criteria

We included randomised controlled trials (RCTs) that compared fertility‐sparing therapy for presumed stage IA grade 1 endometrioid endometrial cancer or AEH with oral progestin compared to levonorgestrel IUS or metformin or other pharmacological interventions, or bariatric or hysteroscopic surgery (any comparison); or any of these interventions with the usual treatment (surgery). Other comparative non‐randomised studies were also eligible for inclusion (quasi‐randomised trials, non‐randomised studies (NRS), and prospective and retrospective cohort studies).

Data collection and analysis

Two review authors independently extracted data and assessed the methodological quality of the studies. We used standard Cochrane methodological procedures. Where possible, we pooled data from RCTs in a meta‐analysis. Otherwise, we provided a narrative description of the results. Primary outcomes are overall survival and live birth rate. Secondary outcomes are progression‐free survival, complete pathological response rate (CR), severe adverse events, psychological symptoms, quality of life, pregnancy rate, and surgery for persistent/progressive disease. We assessed the certainty of the evidence using GRADE. We assessed the risk of bias only in RCTs, using the Cochrane risk of bias tool, RoB 1.

Main results

We included 12 studies with 904 participants; six RCTs and six NRSs. Four studies included women with AEH, two, women with endometrial cancer, and six, both AEH and endometrial cancer. We judged the studies at high risk of overall bias. We pooled two RCTs into one meta‐analysis and described the remaining comparisons narratively.

None of the included studies provided evidence for overall survival, progression‐free survival or quality of life for any comparison.

Metformin plus progestin compared with progestin

Metformin plus progestin may have little to no effect on live birth rate (risk ratio (RR) 1.80, 95% confidence interval (CI) 0.88 to 3.68); 2 RCTs, 72 women; low‐certainty evidence) but may slightly increase CR (RR 1.85, 95% CI 1.07 to 3.19; P = 0.03; 2 RCTs; 141 women; low‐certainty evidence).

No fatal adverse events were observed. Weight gain was the most frequent adverse event in one RCT, with 5/74 (6.8%) cases of grade 3‐4 weight gain in the progestin group versus 2/76 (2.6%) in the metformin plus progestin group (RR 0.39, 95% CI 0.08 to 1.95; 1 RCT; 150 women; low‐certainty evidence). Metformin plus progestin may make little to no difference in the need for surgery for persistent/progressive disease (RR 0.96, 95% CI 0.24 to 3.78; 2 RCTs; 166 women; low‐certainty evidence).

Levonorgestrel IUS compared to oral progestin

Only one RCT evaluated live birth rate, showing little to no difference between levonorgestrel IUS and oral progestin (RR 1.80, 95% CI 0.74 to 4.39; 1 RCT, 34 women; low‐certainty evidence). Data from two RCTs showed no evidence of a difference in CR in women with AEH (data not pooled): RR 1.78 (95% CI 0.98 to 3.25; 89 women), and RR 1.24 (95% CI 0.86 to 1.78; 19 women), both low‐certainty evidence. One RCT found that levonorgestrel IUS may decrease severe adverse events (weight gain) slightly (RR 0.19; 95% CI 0.04 to 0.84; 1 RCT, 118 women; low‐certainty evidence). Evidence on surgery for persistent/progressive disease information was incomplete.

Oral progestin plus levonorgestrel IUS compared to oral progestin

One RCT in endometrial cancer evaluated live birth rate, finding no difference between oral progestin plus levonorgestrel IUS and oral progestin alone (RR 1.40, 95% 0.46 to 4.24; 1 RCT, 33 women; low‐certainty evidence). Similarly, no differences were found in women with AEH (RR 1.38, 95% CI 0.50 to 3.82;1 RCT, 47 women; low‐certainty evidence).

Data from two RTCs showed no difference in CR in women with endometrial cancer (RR 1.30, 95% CI 0.47 to 3.59; 1 RCT, 54 women) or with AEH (RR 1.45, 95% CI 0.77 to 2.76; 1 RCT, 86 women low‐certainty). The only grade 3 adverse event was weight gain, with no difference between the groups for endometrial cancer (RR 1.11, 95% CI 0.17 to 7.34; 1 RCT, 59 women; low‐certainty evidence) and AEH (RR 0.97, 95% CI 0.43 to 2.20; 1 RCT, 112 women; low‐certainty evidence). Surgery for persistent/progressive disease was also similar in the two treatment arms, both for women with endometrial cancer (RR 2.07, 95% CI 0.20 to 21.60; 1 RCT, 59 women; low‐certainty evidence) and with AEH (RR 1.00, 95% CI 0.06 to 15.48; 1 RCT, 86 women; low‐certainty evidence).

Authors' conclusions

In light of the low certainty of the evidence, it is unclear which intervention and which route of administration or dose of progestins could be of benefit compared to others for fertility‐sparing management of endometrial cancer or AEH. The addition of metformin to progestins may increase complete response slightly. Levonorgestrel IUS may result in no difference in efficacy in complete response, compared to oral progestins, whilst it may reduce adverse events slightly. Therefore, levonorgestrel IUS may improve quality of life and compliance with treatment.

Plain language summary

What are the treatment options for women with low‐risk cancer and pre‐cancer of the womb lining (endometrial cancer and atypical endometrial hyperplasia) who wish to have children?

Key messages

• We found many different treatment options for fertility‐sparing treatment of endometrial cancer (cancer of the womb lining) and atypical endometrial hyperplasia (precancer). We don't know which treatment is most effective or what is the best way to deliver it.

• Adding metformin to progesterone treatment could be of benefit when treating endometrial cancer and precancer. Levonorgestrel‐intrauterine system (hormone‐releasing coil) may be of benefit because it obtains a similar response with fewer unwanted effects compared to oral progesterone.

• We need more well‐designed studies to evaluate and compare the different treatment options.

What are endometrial cancer and atypical endometrial hyperplasia?

Endometrial cancer is cancer of the womb lining. It is the sixth most common cancer worldwide and the fourth most common in high‐income countries, where rates are increasing. Atypical endometrial hyperplasia (precancer) is the growth of abnormal cells in the lining of the womb. If it's not treated, it may develop into endometrial cancer. Delays in childbearing and increasing obesity rates are both risk factors for cancer and precancer. The most effective treatment for endometrial cancer and precancer is to remove the womb (hysterectomy). Endometrial cancer and precancer are more common after the menopause, but up to 25% of cases occur in younger women. This means that more women are now diagnosed with cancer whilst still wishing to conceive. This leads to an increasing number of women who may want to consider fertility‐sparing (preserving) treatment.

Treatment with progestins (a female hormone) is usually recommended. In addition, other treatments, such as metformin (a treatment used for diabetes) or bariatric surgery (surgery for obesity) or hysteroscopic resection (a procedure with a thin camera and special tools inserted through the vagina to remove abnormal tissue) are currently being investigated. However, the best treatment options remain uncertain.

What did we want to find out?

We wanted to find out which treatments are effective for endometrial cancer and precancer, and whether they are also effective for preserving women's fertility.

What did we do?

We searched for studies that investigated treatments for endometrial cancer and precancer that also aimed to preserve fertility. We were interested in various treatments, such as different medicines, surgery, and weight‐loss programmes. We wanted to know the effects of the different treatments on women's overall survival, how many women had a baby (live birth rate), how long women survived with no cancer recurrence, how many women completely recovered (complete pathological response rate), any severe unwanted effects of the treatments, whether there were any psychological symptoms or impact on quality of life, how many women became pregnant, and how many needed a hysterectomy following unsuccessful treatment. We summarised and compared the results of the studies and rated our confidence in the evidence, based on factors such as study methods and sizes.

What did we find?

We found 12 studies that involved 904 women. These included six studies where women were randomly assigned to treatments and six studies where treatment was not randomly assigned.

Main results

Metformin plus progestin compared to progestin alone may slightly increase the complete response rate (absence of any detectable cancer after treatment) and may have little to no effect on live births or the need for surgery for persistent/progressive disease. The studies reported no serious unwanted effects. No information is available about the effect on how long women survive overall, or survive before their disease re‐grows, or quality of life.

Levonorgestrel intrauterine system is a hormone‐releasing coil inserted inside the womb. Compared to progestins taken by mouth, it may have no effect on complete response (absence of detectable cancer), and fewer unwanted effects such as weight gain. Women may be more likely to continue treatment with this device. There is no available information about other outcomes.

Adding a hormone‐releasing coil to progestin taken by mouth may make no difference to survival, live births, cure rate, quality of life or need for future surgery. It also appears not to increase unwanted effects.

What are the limitations of the evidence?

Our confidence in the evidence is low, for several reasons. Firstly, in some studies, women were not randomly placed into different treatment groups. This means that differences between the groups could be due to differences between women rather than between the treatments. Secondly, some evidence focused on a specific group of women, such as obese women, whereas the question we hoped to address was broader. Finally, some studies were very small. Therefore, further research is likely to change or confirm these results.

How up‐to‐date is the evidence?

The evidence is current to 3 February 2025.

Summary of findings

Summary of findings 1. Metformin plus progestin compared to progestin.

Metformin plus progestin compared to progestin
Patient or population: women with atypical endometrial hyperplasia (AEH) and endometrial cancer (EC) Setting: university hospital or gynecological oncology unit Intervention: metformin plus progestin (megestrol acetate) Comparison: progestin (megestrol acetate)
Outcomes	№ of participants (studies)	Relative effect (95% CI)	Certainty of the evidence (GRADE)	Anticipated absolute effects* (95% CI)		Comments
				Risk with progestin	Risk difference with metformin plus progestin
Overall survival Follow‐up 12 months to (median) 33.4 (26‐44) months	‐	‐	‐	‐	‐	Not reported
Live birth rate Follow‐up 12 months to (median) 33.4 (26‐44) months	72 (2 RCTs)	RR 1.80 (0.88 to 3.68)	⊕⊕⊝⊝ Lowa,b,c,d,e	225 per 1000	180 more per 1000 (27 fewer to 603 more)	Metformin plus progestin (MA) may result in little to no difference in live birth rate after treatment of endometrial cancer and AEH compared to oral progestin (MA).
Progression‐free survival Follow‐up 12 to (median) 33.4 (26‐44) months	‐	‐	‐	‐	‐	Not reported
Complete pathological response rate (12–16 weeks) Follow‐up 12 months to (median) 33.4 (26‐44) months	141 (2 RCTs)	RR 1.85 (1.07 to 3.19)	⊕⊕⊝⊝ Lowa,c,f,g	212 per 1000	180 more per 1000 (15 more to 465 more)	Metformin plus progestin (MA) may result in a slight increase in the complete pathological response compared to oral progestin (MA).
Severe adverse events (Grade 3+) Follow‐up 12 months to (median) 33.4 (26‐44) months			⊕⊕⊝⊝ Lowh,i,j	No fatal adverse events were observed. One RCT reported 5/74 (6.8%) cases of grade 3‐4 weight gain in the MA‐treated group versus 2/76 (2.6%) in the metformin plus MA group, (RR 0.39; 95% CI 0.08 to 1.95; 1 RCT, 150 women). The most frequent adverse effect was weight gain (grade 1‐4), observed in 29/76 women in the metformin plus MA group; mean weight increase 2.5 kg (‐1 to 6). In the MA group, 34/74 cases gained weight (mean 5 kg (0 to 10; P = 0.01). The study also reported grade 3‐4 adverse events: leukocytosis in 2/74 women (2.7%) in the MA group and 1/76 (1.3%) in the metformin plus MA group; hypercoagulable state in 1/74 women (1.4%) in the MA group and 0/76 in the metformin plus MA group; facial oedema in 1/76 women (1.3%) in the metformin + MA group and 0/74 in the MA group. Another pilot study did not report adverse events.
Quality of life Follow‐up 12 months to (median) 33.4 (26‐44) months	‐	‐	‐	‐		Not reported
Surgery for persistent/progressive disease (hysterectomy) Follow‐up 12 months to (median) 33.4 (26‐44) months	166 (2 RCTs)	RR 0.96 (0.24 to 3.78)	⊕⊕⊝⊝ Lowa,c,k,l	49 per 1000	2 fewer per 1000 (37 fewer to 136 more)	Metformin plus progestin (MA) may result in little decrease to no difference in the need for surgery for persistent/progressive disease in EC and AEH compared to oral progestin.
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). AEH: atypical endometrial hyperplasia; CI: confidence interval; EC: endometrial cancer; MA: megestrol acetate; RCT: randomised controlled trial; RR: risk ratio
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

^aYang 2020 is an RCT classified as having unclear risk of bias in the incomplete outcome data domain. It is at high risk only for the outcome complete pathological response rate (16 weeks), and low risk of bias in the remaining domains. The 'high risk' judgement is applicable to the complete pathological response rate at 16 weeks; not applicable to the other outcomes. For this reason, our judgement on this outcome is 'unclear risk of bias'. We did not downgrade the certainty of evidence because most information is from results at low or unclear risk of bias (some concerns). We rated the study as having no serious limitations. We did not downgrade for risk of bias.
Shan 2014 is a pilot RCT with a small sample size. It is classified as having a high risk of bias in two domains, incomplete outcome data and other sources of bias (it included a subpopulation with metabolic syndrome), and unclear risk of bias in two domains, random sequence generation and allocation concealment. We did not downgrade the certainty of evidence because most information is from results at low or unclear risk of bias (some concerns). We rated the study as having no serious limitations. We did not downgrade for risk of bias.
^bI² = 0%; Chi² = 0.55; P = 0.46. The CI of both studies overlap. No heterogeneity was observed. We did not downgrade because the proportion of the variability in effect estimates that is due to true heterogeneity rather than chance is not important (I² = 0%). We did not downgrade for inconsistency.
^cNo indirectness was observed. Both studies addressed the PICO question and made direct comparisons. No publication bias was suspected. We did not downgrade for indirectness nor publication bias. 
^dThe 95% CI is wide and includes appreciable benefit or harm. The studies include few participants and events. The point estimate reflects a benefit and the boundary of the CI least favourable to the intervention includes the possibility of harm. The ratio of the upper to the lower boundary of the CI is > 3 for risk ratio (ratio = 4.18). We downgraded two levels due to imprecision.
^eShan 2014 did not report data in their study, but information on live birth rate was provided after contacting the authors.
^fI² = 0%; Chi² = 0.69; P = 0.41. The CI of both studies overlap. Heterogeneity was not observed. We did not downgrade because the proportion of the variability in effect estimates that is due to true heterogeneity rather than chance is not important (I² = 0%). We did not downgrade due to inconsistency.
^kI² = 0%; Chi² = 0.56; P = 0.46. The CI of both studies overlap. No heterogeneity was observed. We did not downgrade because the proportion of the variability in effect estimates that is due to true heterogeneity rather than chance is not important (I² = 0%). We did not downgrade due to inconsistency.
^gThe studies include few participants and events. The point estimate reflects a benefit. The boundaries of the CI do not include the possibility of harm. The ratio of the upper to the lower boundary of the CI is < 3 for the risk ratio (ratio = 2.98). We calculated Optimal Information Size (OIS): 102 women in each group (204 women). We downgraded two levels due to imprecision.
^lThe studies include few participants and events. The point estimate reflects a small benefit (less need for surgery) and the boundary of the CI least favourable to the intervention includes the possibility of harm. The ratio of the upper to the lower boundary of the CI is > 3 for the risk ratio (ratio = 15.75). We downgraded two levels due to imprecision.
^hYang 2020 is an RCT classified as having unclear risk of bias in the incomplete outcome data domain, and low risk of bias in the remaining domains. It is at high risk only for the outcome complete pathological response rate (16 weeks). The 'high risk' judgement is applicable for complete response at 16 weeks; not applicable to the other outcomes. For this reason, our judgement on this outcome is 'unclear risk of bias'. We did not downgrade the certainty of evidence because most information is from results at low or unclear risk of bias (some concerns). We rated the study as having no serious limitations. We did not downgrade for risk of bias.
ⁱThe study included few participants and the point estimate reflects a small benefit (less Grade 3‐4 weight gain) and the boundary of the CI least favourable to the intervention includes the possibility of harm. The ratio of the upper to the lower boundary of the CI is > 3 for the risk ratio (ratio = 24.3). We downgraded two levels due to imprecision.
^jNo indirectness was observed. The study addressed the PICO question. The study makes direct comparisons. No publication bias was suspected. We did not downgrade for indirectness nor publication bias.

Summary of findings 2. Levonorgestrel intrauterine system (LNG‐IUS) compared to oral progestin.

Patient or population: women with atypical endometrial hyperplasia (AEH) Setting: university hospital or gynecological centres Intervention: levonorgestrel intrauterine system (LNG‐IUS) Comparison: oral progestin
Outcomes	№ of participants (studies)	Certainty of the evidence (GRADE)	Relative effect (95% CI)	Anticipated absolute effects		Comments
				Risk with oral progestin	Risk difference with LNG‐IUS
Overall survival Not measured	‐	‐	‐
Live birth rate Follow‐up: median 27.8 months	34 (1 RCT)	⨁⨁◯◯ Lowa,b,c,d	In one RCT in women with AEH, after complete pathological response, there were 8/16 (50%) live births in the LNG‐IUS group and 5/18 (27.8%) in the oral progestin (MA) group (RR 1.80, 0.74 to 4.39;34 women; 222 more per 1000, 95% CI from 72 women fewer to 942 more per 1000).			LNG‐IUS may result in little to no difference in the live birth rate compared to oral progestin
Progression‐free survival	‐	‐	‐			Not measured
Complete pathological response rate Follow‐up: median 27.6 months; range 6‐12 months	19 and 120 (2 RCTs)	⨁⨁◯◯ Lowa,b,e,f	Orbo 2014, an RCT with 19 AEH cases, showed a RR 1.24 (95% CI 0.86 to 1.78); 231 more complete pathological response cases per 1000 with LNG‐IUS compared to oral progestin (95% CI from 23 fewer to 577 more) Xu 2023 B), an RCT with 120 participants with AEH, 89 with complete response, showed an RR 1.78(95% CI 0.98 to 3.25); 200 more complete pathological responses per 1000 with LNG‐IUS compared to oral progestin (95% CI from 5 fewer to 576 more)			LNG‐IUS may result in little to no difference in CR compared to oral progestin.
Severe adverse events: grade 3weight gain Follow‐up: median 27.8 months; range 6‐12 months	118 (1 RCT)	⨁⨁◯◯ Lowa,b,c,g	In one RCT with 118 participants with AEH, women had less grade 3 weight gain in the LNG‐IUS group (2/60; 3.3%) compared to the MA group (10/58; 17.2%); RR 0.19 (95% CI 0.04 to 0.84; 140 cases fewer per 1000 (95% CI from 166 fewer to 28 fewer)			LNG‐IUS may slightly reduce grade 3 weight gain compared to oral progestin.
Quality of life	‐	‐	‐			Not measured
Surgery for persistent/progressive disease (hysterectomy)	120 and 19 (2 RCTs)	‐	One RCT with 120 women reported that two women underwent hysterectomy but did not say from which group. Another RCT with 19 women did not report data on surgery for persistent/progressive disease.
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). AEH: atypical endometrial hyperplasia; CI: confidence interval; CR: complete pathological response rate; LNG‐IUS: levonorgestrel intrauterine system; MA: megestrol acetate; RCT: randomised controlled trial; RR: risk ratio
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

^aOrbo 2014 is an RCT classified as being at 'high risk' for other sources of bias. The study included all types of endometrial hyperplasia and had few cases of atypical endometrial hyperplasia. Also, it included 6% of postmenopausal patients, and it does not specify fertility desire. We rated the remaining domains as 'low risk of bias. Allocation was concealed by central telephone randomisation. The research team and pathologists were blinded. Probably participants and gynaecologists were not blinded. However, the authors consider that the outcome is not likely to be influenced by lack of blinding. No incomplete outcome data (10% missing cases) and selective reporting issues were found. We did not downgrade the certainty of evidence because most information is from results at low risk of bias. We rated the study as having no serious limitations. We did not downgrade for risk of bias.
In Xu 2023 B we labelled one domain as 'high risk' and one as 'unclear risk'. The study was at 'unclear risk' of allocation concealment 'high risk' for other sources of bias. The treatment allocation was concealed before enrolment but the concealment method was not specified. We judged the study to have other sources of bias because all the participants underwent additional treatments (hysteroscopic evaluation and resection of lesions). This additional treatment may have concealed the effect of the intervention in both groups. No incomplete outcome data and selective reporting issues were found. We did not downgrade the certainty of evidence because most information is from results at low or unclear risk of bias. We rated the study as having no serious limitations. We did not downgrade for risk of bias.
^bNo inconsistency (one study, Orbo 2014. Studies not combined.
^cNo indirectness was observed. Both studies addressed the PICO question and made direct comparisons. No publication bias was suspected. We did not downgrade for indirectness nor publication bias. 
^dThe study includes a few participants and events. It has a wide CI. The 95% CI includes appreciable benefit or harm. The point estimate reflects a benefit and the boundary of the CI least favourable to the intervention includes the possibility of harm. The ratio of the upper to the lower boundary of the CI is > 3 for the risk ratio (ratio = 5.93). We downgraded two levels due to imprecision.
^eNo inconsistency (one study, Xu 2023 B). Studies not combined.
^fIn one study, Xu 2023 B the point estimate reflects a benefit and the boundary of the CI least favourable to the intervention includes the possibility of harm. The ratio of the upper to the lower boundary of the CI is more than three for the risk ratio. Ratio = 3.31. We downgraded two levels due to imprecision.
Another study, Orbo 2014, includes a few participants and events. The 95% CI includes appreciable benefit; it had a ratio of 2.06. We calculated Optimal Information Size (OIS), and the result was 19 participants per group. We downgraded two levels due to imprecision.
^gThe point estimate reflects a benefit and the boundary of the CI least favourable to the intervention includes the possibility of harm. The 95% CI includes appreciable benefit (RR under 0.75). The ratio of the upper to the lower boundary of the CI is more than three for the risk ratio. Ratio = 21. We downgraded two levels due to imprecision.

Summary of findings 3. Oral progestin plus levonorgestrel intrauterine system (LNG‐IUS) compared to oral progestin.

Patient or population: women with atypical endometrial hyperplasia (AEH) or endometrial cancer (EC) Setting: Obstetrics and Gynaecology Department in university hospital Intervention: oral progestin plus levonorgestrel intrauterine system (LNG‐IUS) Comparison: oral progestin (megestrol acetate (MA))
Outcomes	№ of participants (studies)	Certainty of the evidence (GRADE)	Relative effect (95% CI)	Anticipated absolute effects		Comments
				Risk with oral progestin	Risk difference with oral progestin plus LNG‐IUS
Overall survival	‐	‐	‐			Not measured
Live birth rate Follow‐up: median 31.6 months	(2 RCTs) 56	⨁⨁◯◯ Lowa,b,c,d	Xu 2023 A, an RCT in women with EC showed that 33/56 women who achieved CR planned for parenthood: 5/21 women (23.8%) in the oral progestin (MA) group had a live birth; 4/12 women (33.3%) in the MA plus LNG‐IUS group had a live birth (RR 1.40, 95% CI 0.46 to 4.24; 95 more live births per 1000, 95% CI from 129 fewer to 771 more); 33 women			Oral progestin plus LNG‐IUS may result in little to no difference in live birth rate compared to oral progestin.
Follow‐up: median: 27.8 months	89		Xu 2023 B, an RCT in women with AEH treated 120 women with MA or MA plus LNG‐IUS. After achieving CR, 47 women planned for parenthood: 5/18 women (27.8%) in the MA group had a live birth; 5 /13 (38.5%) in the MA plus LNG‐IUS group had a live birth (RR 1.38, 95% CI 0.50 to 3.82; 106 more live births per 1000, 95% CI from 139 fewer to 783 more); 31 women
Progression‐free survival	‐	‐	‐			Not reported
Complete pathological response rate Follow‐up: median 31.6 months	(2 RCTs) 63	⨁⨁◯◯ Lowa,b,c,e	Xu 2023 A, an RCT in 63 women with EC, showed that 5/26 women (19.2%) in the MA group had CR at 16 weeks and 7/28 (25.0 %) in the MA plus LNG‐IUS (RR 1.30, 95% CI 0.47 to 3.59; 58 more CR cases per 1000, 95% CI from 102 fewer to 498 more); 54 women			Oral progestin plus LNG‐IUS may result in little to no difference in complete pathological response at 16 weeks compared to oral progestin.
Follow‐up: median 27.8 months	120		Xu 2023 B, an RCT in 120 women with AEH, showed that 11/43 women (25.6%) achieved CR at 16 weeks in the MA group and 16/43 (37.2%) in the MA plus LNG‐IUS group (RR 1.45, 95% CI 0.77 to 2.76; 115 more CR cases per 1000, 95% CI 59 fewer to 450 more) 86 women
Severe adverse events: grade 3 weight gain Follow‐up: median 31.6 months	(2 RCTs) 63	⨁⨁◯◯ Lowa,b,c,f	In Xu 2023 A, an RCT in women with EC, the only grade 3 adverse event was weight gain: 2/31 cases (6.5%) in the MA group and 2/28 cases (7.1%) in the MA plus LNG‐IUS group (RR 1.11, 95% CI 0.17 to 7.34;7 more weight gain cases per 1000, 95% CI from 54 fewer to 409 more) 59 women			Oral progestin plus LNG‐IUS may result in little to no difference in severe adverse effects grade 3 (weight gain) compared to oral progestin. No treatment‐related deaths or serious adverse events (grade 4) were observed.
Follow‐up: median 27.8 months	120		In Xu 2023 B, an RCT in women with AEH, the only grade 3 adverse event was weight gain: 10/58 women (17.2%) in the MA group and 9/54 (16.7%) in the MA plus LNG‐IUS grouup (RR 0.97, 95% CI 0.43 to 2.20; 5 fewer weight gain cases per 1000, 95% CI from 98 fewer to 207 more) 112 women
Quality of life	‐	‐	‐			Not measured
Surgery for persistent/progressive disease (hysterectomy) Follow‐up: median 31.6 months	(2 RCTs) 63	⨁⨁◯◯ Lowa,b,c,g	In Xu 2023 A, an RCT in women with EC, 1/30 women (3.3%) in the MA group underwent surgery and 2/29 (6.9%) in the MA plus LNG‐IUS group (RR 2.07, 95% CI 0.20 to 21.60. 36 more women per 1000 needing surgery, 95% CI from 27 fewer to 687 more) 59 women			Oral progestin plus LNG‐IUS may result in little to no difference in surgery for persistent/progressive disease compared to oral progestin.
Follow‐up: median 27.8 months	120		In Xu 2023 B, an RCT in women with AEH, 1/43 women (2.3%) in the MA group underwent surgery and 1/43 (2.3%) in the MA plus LNG‐IUS group (RR 1.00, 95% CI 0.06 to 15.48; 0 fewer women per 1000 needing surgery, 95% CI from 22 fewer to 335 more) 86 women
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). AEH: atypical endometrial hyperplasia; CI: confidence interval; CR: complete pathological response; EC: endometrial cancer; LNG‐IUS: levonrgestrel intrauterine system; MA: megestrol acetate; RR: risk ratio
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

^aIn we labelled one domain as high risk and one as unclear risk. We labelled this study as being at unclear risk of bias for allocation concealment and high risk for other sources of bias. The treatment allocation was concealed before enrolment but the method used was not specified. We judged the study to be at risk of other sources of bias because all the participants underwent additional treatments (hysteroscopic evaluation and resection of lesions). This additional treatment may have concealed the effect of the intervention in both groups. No incomplete outcome data and selective reporting issues were found. We did not downgrade the certainty of evidence because most information is from results at low or unclear risk of bias. We rated the study as having no serious limitations. We did not downgrade for risk of bias.was classified as being at high risk of bias: one domain was labelled as high risk, and one as unclear risk. We rated the study at unclear risk of bias for allocation concealment because the study authors did not specify if allocation was concealed. We judged the study to have other sources of bias because all the participants underwent additional treatments (hysteroscopic evaluation and resection of lesions). This additional treatment may have concealed the effect of the intervention in both groups. No incomplete outcome data or selective reporting issues were found. We did not downgrade the certainty of evidence because most information is from results at low or unclear risk of bias. We rated the study as having no serious limitations. We did not downgrade for risk of bias.
^bNo Inconsistency (studies not combined)
^cNo indirectness was observed. Studies address the PICO question and make direct comparisons. We did not downgrade for indirectness. 
^dXu 2023 A includes few participants and events. It has a wide CI. The point estimate reflects a benefit and the boundary of the CI least favourable to the intervention includes the possibility of harm. The ratio (for EC) of the upper to the lower boundary of the CI is > 3 for the risk ratio (ratio for EC = 9.21). We downgraded two levels due to imprecision. 
Xu 2023 B: the point estimate reflects a benefit and the boundary of the CI least favourable to the intervention includes the possibility of harm. The ratio (for AEH) of the upper to the lower boundary of the CI is > 3 for the risk ratio (ratio for AEH = 7.64). We downgraded two levels due to imprecision.
^eXu 2023 AThe point estimate reflects a benefit and the boundary of the CI least favourable to the intervention includes the possibility of harm. The ratio (for EC) of the upper to the lower boundary of the CI is >3 for the risk ratio (ratio for EC = 7. 63). We downgraded two levels due to imprecision. 
Xu 2023 B The point estimate reflects a benefit and the boundary of the CI least favourable to the intervention includes the possibility of harm. The ratio (for AEH) of the upper to the lower boundary of the CI is > 3 for the risk ratio (ratio for AEH = 3.58). We downgraded two levels due to imprecision.
^fXu 2023 AThe point estimate reflects a benefit and the boundary of the CI least favourable to the intervention includes the possibility of harm. The ratio (for EC) of the upper to the lower boundary of the CI is > 3 for the risk ratio (ratio for EC = 43.17). We downgraded two levels due to imprecision. 
Xu 2023 B The point estimate reflects a benefit and the boundary of the CI least favourable to the intervention includes the possibility of harm. The ratio (for AEH) of the upper to the lower boundary of the CI is > 3 for the risk ratio (ratio for AEH = 5.11). We downgraded two levels due to imprecision. 
^gXu 2023 AThe point estimate reflects a benefit and the boundary of the CI least favourable to the intervention includes the possibility of harm. The ratio (for EC) of the upper to the lower boundary of the CI is > 3 for the risk ratio (ratio for EC = 108). We downgraded two levels due to imprecision.
Xu 2023 B The point estimate reflects a benefit and the boundary of the CI least favourable to the intervention includes the possibility of harm. The ratio (for AEH) of the upper to the lower boundary of the CI is > 3 for the risk ratio (ratio for AEH = 258). We downgraded two levels due to imprecision.

Background

Description of the condition

Endometrial cancer (cancer of the lining of the womb) is the sixth most common cancer in women worldwide (Sung 2021), and the fourth most common cancer in women in high‐income countries, where its incidence is increasing (Cancer Research 2025; Crosbie 2022; SEER 2022). Each year, approximately 417,000 new cases of endometrial cancer are diagnosed worldwide: 9800 in the UK and 67,880 in the USA (Cancer Research 2025; SEER 2022; Siegel 2022; Sung 2021).

Endometrial hyperplasia is an overgrowth of the womb lining and can be a precursor to some types of endometrial cancer. It is a proliferative endometrial lesion. The incidence of endometrial hyperplasia is 133 per 100,000 women per year, and that of atypical endometrial hyperplasia (AEH) is 54 per 100,000 women per year (Reed 2009).

The 2014 World Health Organization (WHO) classification of endometrial hyperplasia includes only two categories: hyperplasia without atypia, and atypical hyperplasia (AEH)/endometrial intraepithelial neoplasia (Zaino 2014). This reduction to just two categories includes new findings on molecular genetic changes. Hyperplasia without atypia has no significant genetic changes and usually regresses. Atypical endometrial hyperplasia shares many mutations with endometrioid endometrial cancer (such as microsatellite instability, paired box gene2 (PAX2) inactivation, phosphatase and tensin homolog (PTEN), KRAS, CTNNB1 (β‐catenin) mutation), and has a high co‐incidence or conversion to endometrial cancer Zaino 2014.

Hyperplasia with and without atypia have different rates of progression to cancer. Hyperplasia without atypia has a reported progression rate of 0.6% of cases (1% to 5%) (Kurman 1985; Lacey 2010; Ordi 2014), whereas atypical hyperplasia progresses in as many as 8% of cases (19% to 27%), and this may be an underestimate, since co‐existing carcinoma has been reported in 36% to 59% of women undergoing hysterectomy (womb‐removing surgery) for AEH (Antonsen 2012; Kurman 1985; Lacey 2010; Rakha 2012; Zaino 2014).

Endometrial cancer is most common in postmenopausal women. However, between 14% and 25% of cases are diagnosed in premenopausal women (Barakat 2006; Creasman 2001; Lee 2007; Obermair 2020; Rodolakis 2015; Siegel 2015), with 4% to 5% of cases in women aged under 40 years, and 7% to 10% in women aged under 45 years (Barakat 2006; Creasman 2001; Lee 2007; Navarria 2009; Rodolakis 2015; SEER 2022).

Endometrioid adenocarcinoma of the endometrium is the most common type of endometrial cancer. It is often diagnosed at an early stage, which usually has a very good prognosis (ACS 2024). In young women, five‐ and 10‐year disease‐free survival after standard surgical treatment (hysterectomy and bilateral salpingo‐oophorectomy) is 99.2% and 98%, respectively (Crissman 1981; Lajer 2012). It is commonly thought that young women with endometrial cancer are more likely to have early‐stage, low‐risk tumours. However, in a population‐based registry, only 18% of women aged younger than 45 years had stage IA grade 1 endometrial cancer (Navarria 2009).

The 2023 revised FIGO (International Federation of Gynaecology and Obstetrics (Fédération Internationale de Gynécologie et d’Obstétrique)) staging system introduced significant changes to Stage I (Berek 2023).

In most instances, stage I is limited to tumours that are confined to the uterine corpus, characterised by non‐aggressive histological types (such as low‐grade endometrioid carcinoma), the absence of substantial or extensive lymphovascular space invasion (LVSI), or aggressive histological types that do not exhibit myometrial invasion.

The updated endometrial cancer staging system incorporates the following changes for stage I (Berek 2023).

• IA1: non‐aggressive histological types confined to an endometrial polyp or limited to the endometrium

• IA2: non‐aggressive histological types involving less than half of the myometrium, with no or focal LVSI

• IA3: low‐grade endometrioid carcinomas confined to the uterus, with concurrent low‐grade endometrioid ovarian involvement

In 2013, a new molecular classification system defined by The Cancer Genome Atlas (TCGA) identified four prognostic categories for endometrial cancer (Cancer Genome Atlas Research Network 2013). Several research groups have extrapolated these findings from TCGA utilising clinically applicable methods. One approach is ProMisE (proactive molecular risk classifier for endometrial carcinoma). It uses a simplified surrogate that includes three immunohistochemical markers (p53, MSH6, and PMS2) and one molecular test (study of pathogenic polymerase epsilon (POLE) mutations) to identify endometrial cancer with mismatch repair deficiency (MMRd), mutations in the exonuclease domain of DNA POLE, and wild type or aberrant p53 expression (p53 wt or p53 abn respectively) (Talhouk 2017). ProMisE has identified four molecular subtypes of endometrial cancer that are similar, yet not identical, to the four genomic subtypes outlined in TCGA (Makker 2022). Integrating molecular classification into existing risk stratification models for endometrial cancer significantly improves prognostic prediction compared to using either system independently (Crosbie 2022).

The main risk factor for endometrial hyperplasia and endometrioid endometrial adenocarcinoma is obesity. Obesity is associated with the peripheral conversion of androgens to oestrogens by adipose tissue (Chen 2017). Other risk factors include a sedentary lifestyle, hyperinsulinaemia (excess insulin), diabetes mellitus, hypercholesterolaemia (high cholesterol), hypertension (high blood pressure), nulliparity (not having had any children), early menarche (a first period at an early age) and anovulation (lack of ovulation or irregular periods) (Corzo 2018), but many of these factors are related to, and not independent of, obesity. Polycystic ovarian syndrome (PCOS) is the most common endocrine disorder associated with anovulation (Gottschau 2015), and it is also a common cause of infertility. Therefore, in women with PCOS, there may be an increase in peripheral conversion of oestrogens and also the associated anovulation results in lack of progestin release. Progestin is produced in ovulatory cycles by the ovarian corpus luteum. Furthermore, nulliparity itself is a risk factor for endometrial cancer and is inversely related to parity, since nulliparous women miss out on periods of high progestin levels during a pregnancy (Johnson 2023; Wu 2015; Yang 2015).

Some premenopausal women with endometrial cancer may be interested in fertility‐sparing treatment, especially those who have not completed their families. These women should be informed that standard surgical treatment for endometrial cancer/atypical hyperplasia has excellent survival results, but will mean they are no longer able to carry a pregnancy. Unfortunately, many of these women will have underlying subfertility, so, even without a diagnosis of atypical hyperplasia or endometrial cancer, the chance of becoming pregnant may be low. Fertility after conservative treatment for endometrial atypical hyperplasia and cancer will depend on the response to cancer treatment and also on underlying patient factors affecting fertility (e.g. presence of PCOS). Discussion about the appropriateness of fertility‐sparing surgery should therefore balance potential risks with a realistic discussion about the chance of achieving a pregnancy on an individual basis. Ovulation induction treatment may be necessary (Jin 2018), which may be associated with an increased risk of endometrial atypical hyperplasia and cancer (Skalkidou 2017). In relation to fertility after treatment with progestins, some studies have shown a live birth rate between 25% and 35%; this being higher with assisted reproductive techniques (Erkanli 2010; Gallos 2012).

Treatment of endometrial atypical hyperplasia and endometrial cancer with progestins may also be considered for (usually) older, postmenopausal women who are at very high risk of surgical treatment. However, this is a different clinical scenario, with different factors driving decision‐making. Treatment of women who are at very high risk for surgical treatment is therefore outside the scope of this Cochrane review.

Description of the intervention

The European Society of Gynaecological Oncology (ESGO)/European Society of Human Reproduction and Embryology (ESHRE)/European Society for Gynaecological Endoscopy (ESGE) guidelines recommend that fertility‐sparing treatment is considered for women with endometrioid endometrial carcinoma Grade 1, Stage IA, without myometrial invasion and without risk factors (Level of evidence V, Grade A; Rodolakis 2023). The guidelines recommend establishing a reliable histopathology with hysteroscopic‐guided endometrial biopsy; preoperative assessment of myometrial invasion should be performed using magnetic resonance imaging (MRI) or transvaginal ultrasound by a specialised radiologist or sonographer (Level of evidence III, Grade A); adnexal involvement should be ruled out by pelvic MRI or transvaginal ultrasound (Level of evidence II, B; Rodolakis 2023).

The 2025 National Comprehensive Cancer Network (NCCN) includes the following criteria to consider fertility‐sparing management of endometrial cancer: well‐differentiated (grade 1) endometrioid adenocarcinoma on dilation and curettage confirmed by expert pathology review; disease limited to the endometrium on MRI (preferred) or transvaginal ultrasound; absence of suspicious or metastatic disease on imaging; no contraindications to medical therapy or pregnancy. Also, NCCN recommends molecular evaluation of tumours and evaluation of inherited cancer risk (Abu‐Rustum 2025).

In young women with low‐grade, Stage IA endometrial cancer, the most significant benefit from progestin treatment is observed in those with p53 wild‐type molecular type tumours. Other molecular types are less likely to respond to progestins. Therefore, the ESGO/ESHRE/ESGE guidelines recommend the use of the ProMisE molecular classifier for all young women with Grade 1, low‐stage endometrial cancer who want fertility‐sparing treatment, although current available data do not support its clinical applicability (Level of evidence IV, Grade B; Rodolakis 2023). An immunohistochemistry study to identify mismatch repair‐deficient tumours is essential for detecting patients at high risk for Lynch syndrome (Level of evidence III, Grade A).Rodolakis 2023. Other scientific societies establish similar recommendations to select candidates for fertility‐sparing management (Burke 2014; Morrison 2022; Rodolakis 2015).

Although FIGO staging is based on histological examination of the surgical specimen (Pecorelli 2009), in women not undergoing hysterectomy staging relies on inference from imaging and from clinical staging. Contrast‐enhanced MRI is the best imaging test for detecting myometrial, cervical invasion or lymph node metastasis compared with non‐enhanced MRI, computed tomography (CT) or ultrasound (Beddy 2012; Burke 2014; Kitajima 2009; Rodolakis 2015; Selman 2008). MRI can also exclude obvious synchronous ovarian tumours or lymph node metastasis. Recent studies indicate that transvaginal ultrasound and pelvic MRI demonstrate comparable diagnostic performance in evaluating myometrial invasion and cervical stromal invasion in early endometrial carcinoma. The diagnostic accuracy of both transvaginal ultrasound and MRI is maximised when conducted by experienced practitioners (Rodolakis 2023).

Despite this pretreatment evaluation, there is a 15% to 30% risk that the disease is of a higher grade or more widespread than observed by pretreatment assessment (Eltabbakh 2005; Leitao 2009). In endometrial atypical hyperplasia there is a risk of co‐existing ovarian cancer (4%), and a 2% risk of progression to a higher stage than stage I endometrial cancer. Synchronous or metastatic ovarian cancer can occur in 5% to 29% of women with endometrial carcinoma, and women older than 45 years are five times more likely to have synchronous ovarian cancer than women younger than 45 years (Obermair 2020). A transvaginal ultrasound scan or an MRI scan or both is therefore recommended, if conservative treatment is being considered (Gallos 2012; RCOG 2016; Rodolakis 2023).

Pharmacological and non‐pharmacological interventions

Medical management of endometrioid endometrial cancer and atypical hyperplasia has been based on progestins (Corzo 2018; Morrison 2022; Niwa 1997; Plaxe 2016; Rodolakis 2023), given orally or via a levonorgestrel‐releasing intrauterine system (IUS) (Burke 2014; RCOG 2016; Sundar 2017). A few studies and case series have shown the efficacy of metformin in proliferative disorders of the endometrium (Tabrizi 2015), including AEH and early‐stage endometrial cancer (Session 2003). Other hormone treatments acting on oestrogen levels include gonadotropin‐releasing hormone (GnRH) agonists and aromatase inhibitors.

Other pharmacological interventions include drugs to reverse insulin resistance, such as metformin, or pharmacological interventions to promote weight loss, such as appetite suppressants or drugs to reduce fat absorption.

Non‐pharmacological interventions can include ‘lifestyle' interventions aimed at promoting weight loss. Their objective is to reduce nutrient intake (diet) and increase physical activity. These interventions are beyond the scope of this Cochrane review since, alone, they are unlikely to have a rapid enough effect in the presence of an otherwise untreated cancer. Interventions to promote weight loss in women conventionally treated for endometrial cancer have been reviewed elsewhere (Kitson 2018).

Surgical interventions

Hysteroscopy is a procedure performed to examine the endometrial cavity by using a video camera on the tip of a thin lighted tube (hysteroscope) inserted through the cervical canal of the uterus. Hysteroscopic tumour resection could be performed as a treatment itself or as a first procedure amongst different steps in a conservative treatment for endometrial cancer.

The hysteroscopic treatment for focal Stage 1A endometrial cancer is conducted using a three‐step technique.

• Step 1: the tumour is excised

• Step 2: the endometrium surrounding the tumour is removed

• Step 3: the myometrium underlying the tumour is removed.

Additionally, multiple random biopsies are taken (Mazzon 2005; Mazzon 2010). For patients diagnosed with AEH, the hysteroscopic approach involves superficial endometrial resection, preserving the basal layer of the endometrium (Giampaolino 2019).

Bariatric, or weight‐loss surgery, is a treatment for morbid obesity. It is the only treatment for obesity shown to deliver definitive weight loss during long‐term follow‐up (Khwaja 2010). In addition to weight reduction, there is a high rate of improvement or even cure of comorbidities associated with obesity, such as type II diabetes mellitus, obstructive sleep apnoea, hypertension, asthma, osteoarthritis, cancer risk and gastro‐oesophageal reflux disease (Adams 2009; Upala 2015). Techniques include: Roux‐en‐Y gastric bypass; and adjustable gastric band and sleeve gastrectomy, normally via a laparoscopic approach. Bariatric surgery may be carried out within the framework of the primary prevention of endometrial cancer. Data on bariatric surgery, as a treatment for established cancer, are limited at present.

How the intervention might work

Obesity is a major risk factor for endometrial cancer and hyperplasia. A meta‐analysis demonstrates that the risk of endometrial cancer increases as weight increases (Jenabi 2015). In obese women there is a metabolic state that promotes oncogenesis. This metabolic state is related to hyperoestrogenaemia, inflammation and insulin resistance, and leads to multiple changes in oncogenic signalling pathways, providing potential targets for treatment (Mackintosh 2013).

Progestins and hormonal treatment

Common reasons for hyperoestrogenaemia are excessive endogenous oestrogens produced by adipose cells in obese women or chronic anovulation. The goal of hormonal treatments is to counterbalance the action of oestrogens or reduce oestrogen levels. Progestins have been the most widely used hormone treatment, but GnRH agonists or aromatase inhibitors have also been used.

Progestin has an essential role in limiting the proliferative actions of oestrogens on the endometrium. Progestins have been shown to have other anti‐proliferative actions, including: oestrogen receptor inhibition; anti‐angiogenic action; and insulin‐like growth factor 1 (IGF‐1) inhibition (Kim 2013).

Endometrial hyperplasia expresses high levels of progestin receptors (Miyamoto 2004). Low‐grade endometrial cancer often expresses oestrogen and progestin receptors (Darvishian 2004; Demopoulos 1999). Progestin receptor‐negative tumours do not respond to treatment with progestins, although progestin receptor expression alone does not guarantee response to treatment, since less than half of women will demonstrate a sustained response. From studies of progestin treatment for endometrial cancer, complete response rates of 48% to 87% and recurrence rates of 20% to 47% have been reported (Erkanli 2010; Gallos 2012; Gunderson 2012; Ushijima 2007; Wei 2017).

GnRH agonists decrease the concentration of oestrogens by decreasing the release of follicle‐stimulating hormone and Luteinising hormone. Aromatase inhibitors block the production of oestrogens in peripheral adipose tissue.

Reversal of insulin resistance

Insulin resistance and hyperinsulinaemia are key factors in endometrial hyperplasia and can be factors in the initiation of endometrial cancer (Shan 2014). Interventions that work by reversing insulin resistance include pharmacological and non‐pharmacological management (including weight‐loss surgery).

Metformin is an anti‐diabetic drug that regulates glucose metabolism. Metformin has multiple anti‐proliferative mechanisms, including: reduction of insulin and IGF‐1 levels (Ferguson 2013; Kim 2013; Wang 2012); increase in progestin receptor expression; and reduction of progestin resistance (Xie 2011). Results of a meta‐analysis suggest that metformin is useful in the reversal of proliferation biomarkers related to tumour progression, and can lead to the reversal of atypical hyperplasia or improvement in survival of endometrial cancer (Meireles 2017).

Weight loss interventions (lifestyle interventions) in obese women at risk of endometrial cancer produce changes in blood biomarkers associated with endometrial cancer (Yates 2017). Breast cancer can be hormonally‐driven, in the same way as endometrial cancer. In breast cancer, weight‐loss interventions decrease oestradiol levels (total and free). Rock 2013. These interventions, alone or in combination with metformin Patterson 2018 can reduce recurrence and mortality rates.

A meta‐analysis that evaluated the relationship between body mass index (BMI) and mortality in endometrial cancer showed that increased BMI is associated with increased mortality (Secord 2016). Weight‐loss interventions may improve the survival of women with endometrial cancer or atypical hyperplasia through pathways that link obesity and endometrial cancer.

Observational evidence links intentional surgical weight loss with reductions in endometrial cancer incidence (Anveden 2017; Luo 2017). A recent review on bariatric surgery to prevent endometrial cancer development concluded that “bariatric surgery seems to reduce the risk of endometrial cancer” (Winder 2018).

However, there is a dearth of evidence on the effect of weight‐loss interventions on women with established endometrial cancer, as shown in a recent Cochrane review that failed to identify randomised controlled trial (RCT)‐level evidence for assessing the effect of bariatric surgery on women with endometrial cancer following conventional treatment (Kitson 2018).

Bariatric surgery can lower the levels of glucose, insulin and insulin‐like growth factor‐binding protein 1 (IGFBP1) and improve insulin sensitivity in obese women after surgical treatment (Linkov 2017; Modesitt 2015). These surgical weight‐loss interventions are associated with a decrease in the expression of inflammatory endometrial cancer risk biomarkers (Linkov 2017).

Bariatric surgery has proven useful in the prevention of endometrial cancer. At the moment, there is a lack of high‐quality evidence that weight loss is an effective treatment for endometrial cancer or atypical hyperplasia. However, few cases of response to weight‐loss surgery have been reported. A prospective pilot study with a cohort of obese women undergoing bariatric surgery showed a reduction in endometrial pathology (Argenta 2013). A case report of surgical weight‐loss treatment of endometrial cancer observed a complete pathological response (Benito 2015).

Hysteroscopic resection

The combination of hysteroscopic resection and levonorgestrel IUS has been proposed as an alternative to first‐line treatment with progestins alone (Rodolakis 2023). In a retrospective cohort study involving 69 women with endometrial cancer and AEH, the study authors reported that this combination achieved a complete response rate of 78.6% and 92.7%, respectively, along with a live birth rate of 40%. The surgical technique is described in the surgical intervention section (Giampaolino 2019).

Why it is important to do this review

Delays in child‐bearing age and increasing obesity rates mean that more women are now diagnosed with endometrial cancer whilst still wishing to conceive, resulting in an increasing number of women who may therefore wish to explore fertility‐sparing management.

Progestin treatment is recommended by most guidelines as first‐line treatment in endometrial cancer or AEH in women who wish to preserve fertility. However, the data on comparisons of different drugs or doses are variable (Abu‐Rustum 2025; Burke 2014; Morrison 2022; RCOG 2016; Rodolakis 2015; Rodolakis 2023). The drug, route of administration, dose and duration of optimal treatment for these women remains unclear. In addition, other treatments, such as metformin or surgery (bariatric, hysteroscopic resection) or both, are currently being investigated and may have a place in the conservative approach.

Women considering fertility‐sparing treatment could otherwise have an excellent prognosis with standard surgical treatment. Therefore, it is important for women and their caregivers to understand the limitations of effectiveness and increased risks of treatment, and to make informed decisions. These discussions should be in the context of realistic statistics of an individual's chance of achieving a pregnancy.

This Cochrane review aims to clarify the effectiveness and risks of fertility‐sparing therapies, including overall survival, complete pathologic response rate, live birth rates, progression‐free survival and surgery for persistent or progressive disease (hysterectomy).

Objectives

To compare the effectiveness and safety of fertility‐sparing treatments, including pharmacological interventions (e.g. oral progestin, levonorgestrel IUS, metformin) and bariatric or hysteroscopic surgery, for AEH and presumed stage IA grade 1 endometrioid endometrial cancer.

Methods

Criteria for considering studies for this review

Types of studies

Randomised controlled trials (RCTs) that compared fertility‐sparing therapy for presumed stage IA grade 1 endometrioid endometrial cancer or AEH with oral progestin compared with levonorgestrel IUS or metformin or other pharmacological intervention or bariatric or hysteroscopic surgery (any comparison) were eligible for inclusion. We also included RCTs that compared any of the above interventions with usual care (hysterectomy).

The primary goal of most Cochrane reviews is to identify highly reliable evidence, typically derived from randomised trials. However, since Cochrane reviews aim to inform clinical and policy decisions, we also drew on the best available evidence rather than exclusively focusing on the highest tier for our question.

We included non‐randomised studies of interventions (NRSI) because we anticipated that our review question would be challenging to answer solely through a review of RCTs. We believed that there would be either no or insufficient RCTs to adequately address our question of interest.

We did not combine other comparative non‐randomised studies, such as quasi‐randomised trials, non‐randomised trials, and prospective and retrospective cohort studies, in a meta‐analysis because we found RCTs in the search. We summarise non‐randomised studies narratively.

We considered it appropriate to pull together data from women with AEH (pre‐malignant lesion) and stage IA1 endometrial cancer in the same analysis. The two entities share mutations; they have a high rate of coincidence in the same patient, and progression to carcinoma.

Types of participants

• Women 18 years of age or older, who desire fertility conservation, with endometrioid endometrial cancer, FIGO stage IA, grade 1, without evidence of pelvic lymph node metastasis or extrauterine disease on CT imaging or MRI.

• Women 18 years of age or older who desire fertility conservation, with AEH.

Types of interventions

Studies that compare fertility‐sparing treatment based on the following treatment strategies.

• Hormonal manipulations including: progestin treatment (including but not limited to oral and intra‐uterine administration); gonadotrophin‐releasing analogues; aromatase inhibitors

• Reversal of insulin resistance including:

  • pharmacological interventions (such as, but not limited to, metformin, appetite suppressants, drugs that cause fat malabsorption or serotonin receptor antagonists (drugs that affect appetite) of any dose, route of delivery or duration);

  • surgical interventions (including, but not limited to, gastric band, sleeve (surgical removal of part of the stomach), or bypass procedure or hysteroscopic surgery)

Any of these interventions compared with any other intervention or usual care (womb‐removing surgery).

We included studies with cointerventions, as long as they were balanced in all study arms.

Types of outcome measures

Primary outcomes

• Overall survival: time from study enrolment to death from any cause

• Live birth rate: number of live births per 1000 women during a specified period

Secondary outcomes

• Progression‐free survival: time from study enrolment to death from any cause

• Complete pathological response rate: histological regression with a lack of histological evidence of cancer in the endometrium. This includes proliferative, secretory, atrophic or inactive endometrium. We recorded the presence of endometrial hyperplasia, with and without atypia.

• Severe adverse events (grade 3+) classified according to Common Terminology Criteria for Adverse Events (CTCAE) Version 5.0 (CTCAE 2017). For example, anaemia: haemoglobin over 8 g/dL is a grade 3 severe adverse event. Nausea: inadequate oral caloric or fluid intake; tube feeding, total parenteral nutrition, or hospitalisation indicated are grade 3 events.

• Quality of life measured using validated scales (e.g. European Organisation for Research and Treatment of Cancer. Quality of Life Questionnaire C 30 (EORTC QLC C‐30), Short Form Health Survey 36 (SF‐36).

• Psychological symptoms (anxiety, depression), measured using validated scales (e.g. Hospital Anxiety and Depression Scale (HADS scale)).

• Pregnancy rate: number of pregnancies per 1000 women

• Surgery for persistent/progressive disease (hysterectomy)

Search methods for identification of studies

Electronic searches

We searched the following electronic databases on 3 February 2025. Searches were not limited by language, date or status of publication.

• Cochrane Central Register of Controlled Trials (CENTRAL; 2024, Issue 12) in The Cochrane Library

• MEDLINE via Ovid (1946 to 31 January 2025);

• Embase via Ovid ( 1974 to 31 January 2025).

We present the search strategies for MEDLINE, Embase and CENTRAL in Appendix 1.

Searching other resources

As we identified ongoing trials that have not yet been published via the above searches, we contacted the principal investigators for relevant data. We also approached the major co‐operative trial groups active in this area for information.

Handsearching

We checked the citation lists of included studies, key textbooks and systematic reviews through handsearching. We also contacted experts in the field to identify further reports of trials. Handsearching includes work published in the last two years. We searched the conference reports from the following sources.

• Gynecologic Oncology (Meeting of the American Society of Gynecologic Oncologists)

• International Journal of Gynecological Cancer (Meeting of the International Gynecologic Cancer Society)

• British Gynaecological Cancer Society

• British Cancer Research Meeting

• Meeting of the European Society of Medical Oncology (ESMO)

• Meeting of the American Society of Clinical Oncology (ASCO)

• Meeting of the European Society of Human Reproduction (ESHRE)

• Meeting of the American Society of Reproductive Medicine (ASRM)

• Meeting GyneGynaecologiccologic Oncology Group (GOG)

• Meeting of the European Society of Gynaecological Oncology (ESGO)

• Meeting of the International Fertility Preservation Society (IFPS)

• Cancer Research UK website

In addition, we searched the following sources.

• ClinicalTrials.gov (https://clinicaltrials.gov/)

• European Organisation for Research and Treatment of Cancer (EORTC) (www.eortc.org/)

• Southwestern Oncology Group (http://swog.org)

• Medical Research Council Clinical Trials Unit at University College of London (UCL) (http://www.ctu.mrc.ac.uk/)

• National Cancer Institute of America (www.cancer.gov/clinicaltrials)

• National Cancer Institute Canada (https://www.ctg.queensu.ca/public/clinical-trials)

• National Health and Medical Research Council of Australia (www.ctc.usyd.edu.au).

• UK's Clinical Study Registry (ISRCTN).

Furthermore, we searched the reference lists of studies that met the inclusion criteria.

Data collection and analysis

Selection of studies

We downloaded all titles and abstracts retrieved by electronic searching to a reference management database (Covidence). We removed duplicates, and two review authors (MEF, NC, PV, JL and CJ) independently examined the remaining references. We excluded studies that clearly did not meet the inclusion criteria and obtained full‐text copies of potentially relevant references. Two review authors (MEF, NC, PV, CO and CJ) independently assessed the eligibility of retrieved papers and resolved any disagreements by discussion. We consulted another review author (JS) if necessary. We listed the reasons for the exclusion of the full‐text article in the Characteristics of excluded studies. We have illustrated the study selection process in a PRISMA diagram (Figure 1; Moher 2009).

1.

If a study included information on fertility‐sparing treatment and non‐surgical treatment of endometrial cancer in women not candidates for surgery, we tried to contact the study authors to determine which group corresponded to each case.

Data extraction and management

Two review authors (MEF and JS) independently collected data for the review using a form that included the following general information: title, authors, contact direction, year of publication, duplicate publications, context, and financing.

We removed duplicates and excluded studies that clearly did not meet the inclusion criteria. We documented the reasons for exclusion. To obtain unreported data, we contacted the first author of the included studies.

We reached out to the corresponding authors of seven studies to gather or clarify information regarding their research. In one instance, the authors informed us that they were unable to share the results at that time. However, in two cases related to the included studies, the authors were able to provide the requested information. Unfortunately, we did not receive responses from the remaining authors.

• For included studies, two review authors (MEF or JS) and (NC, PV, JL or CJ) independently extracted data on to a data abstraction form specially designed for the review of the characteristics of:

  • women (inclusion criteria, age, FIGO staging and year, grade, histology, histological subtype, depth of myometrial invasion, number enroled in each arm); and

  • interventions (described in the Types of interventions section);

  • risk of bias;

  • duration of follow‐up;

  • outcomes; and

  • deviations from protocol.

• For time‐to‐event (survival and disease progression) data, we extracted the log of the hazard ratio (HR) and its standard error from the study reports; if they were not reported, we attempted to estimate the log (HR) and its standard error using the methods of Parmar 1998.

• For dichotomous outcomes (e.g. adverse events or deaths if it was not possible to use HR), we extracted the number of participants in each treatment arm who experienced the outcome of interest and the number of participants assessed at endpoint, in order to estimate the risk ratio.

• For continuous outcomes (e.g. quality‐of‐life measures), we extracted the final value and standard deviation (SD) of the outcome of interest and the number of participants assessed at endpoint in each treatment arm at the end of follow‐up, in order to estimate the mean difference (MD) between treatment arms and its standard error.

We extracted data on both unadjusted and adjusted statistics, if reported. Where possible, all data extracted were those relevant to an intention‐to‐treat analysis, in which we analysed participants in the groups to which they had been assigned in the study. We recorded the time points at which the study authors collected and reported outcomes. We resolved differences in opinion between review authors by discussion or by consulting a third review author (JS) if necessary.

Assessment of risk of bias in included studies

Randomised controlled trials

We assessed the risk of bias in included RCTs using the Cochrane risk of bias tool (RoB 1) and the following criteria (Higgins 2017).

Was the allocation sequence adequately generated?

• Yes (e.g. participants assigned to treatments on the basis of a computer‐generated random sequence or a table of random numbers)

• No (e.g. participants assigned to treatments on the basis of date of birth, clinic ID‐number or surname, or no attempt to randomise participants)

• Unclear (e.g. not reported, information unavailable).

Was allocation adequately concealed?

• Yes (e.g. where the allocation sequence could not be foretold)

• No (e.g. allocation sequence could be foretold by patients, investigators or treatment providers)

• Unclear (e.g. not reported)

Were women, treatment providers and outcome assessors adequately prevented from knowing the allocated interventions during the study?

• Yes

• No

• Unclear

Was loss to follow‐up less than 20% and were the reasons for loss to follow‐up similar in both arms?

• Yes

• No

• Unclear

Are reports of the study free of suggestion of selective outcome reporting?

• Yes (e.g. if the study reports all outcomes specified in the protocol)

• No

• Unclear

Was the study apparently free of other problems that could put it at a high risk of bias?

• Yes

• No

• Unclear

Two review authors (MEF and JS) independently applied RoB 1 and resolved any differences in opinion by discussion or by appeal to a third review author (NC). We summarised the results in Figure 2 and Figure 3. Our interpretation of the meta‐analysis results took into account the findings related to risk of bias.

2.

'Traffic light' plots of the domain‐level judgements for each individual result

3.

Weighted bar plots of the distribution of risk‐of‐bias judgements within each bias domain

We rated the overall risk of bias of the included RCTs according to the criteria detailed in the Cochrane Handbook of Systematic Reviews of Interventions (Higgins 2017). The overall risk of bias is the least favourable assessment across the domains of bias. If we labelled all the domains 'low‐risk', then we labelled the RCT as low risk; an RCT at high risk in one or more domains was labelled as high risk overall; we judged an RCT to be at unclear risk of bias overall if we had labelled it as unclear risk in one or more domains.

Non‐randomised studies

We planned to assess risk of bias for non‐randomised studies only if we did not find any RCTs. Had we found only non‐randomised studies (cohort comparative trials), we planned to apply the ROBINS‐I tool, which is a specific tool to evaluate non‐randomised studies' quality (Sterne 2016). The ROBINS‐I tool evaluates the risk of bias of each non‐randomised study in seven domains (confounding, selection of participants, classification of interventions, deviations from intended interventions, missing data, measurement of outcomes and selection of reported results) and finally gives an overall risk of bias of every study.

Measures of treatment effect

We used the following measures of the effect of treatment (Higgins 2024).

• For time‐to‐event outcomes (e.g. overall survival), we extracted the HR with its 95% confidence interval (CI).

• For continuous outcomes (e.g. quality‐of‐life scores), we assumed that the study authors used different measurement scales. Therefore, we planned to estimate the standardised mean difference (SMD) and its 95% CI using the pooled data. However, if studies used the same measurement scale, we estimated the MD and its 95% CI. If studies did not report total values but, instead, reported change‐from‐baseline outcomes, we combined these change values with total measurement outcomes by using the (non‐standardised) MD method in Review Manager (RevMan; RevMan 2025). We used subgroups to distinguish between MDs of change scores and MDs of final values, and pooled the subgroups in an overall analysis.

• For dichotomous outcomes, we calculated the effect size as a risk ratio (RR) with its 95% CI.

Unit of analysis issues

We planned to exclude cross‐over studies and cluster‐RCTs, as these study designs would not be appropriate for the comparisons and conditions considered in the review, but none were identified in the bibliographic search. If we had identified studies with multiple treatment arms, we would have considered them as pair‐wise comparisons and followed the recommendations in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2024).

Dealing with missing data

We analysed and documented the reasons for missing data. To obtain missing data, we contacted the study authors. We did not impute missing data when these data were unavailable.

Assessment of heterogeneity

We considered whether the included studies were clinically and methodologically similar enough to calculate a common measurement of the effect (Deeks 2024). We assessed between‐study heterogeneity by visual inspection of forest plots and superimposition of CIs, and also by analysing the populations studied. We performed a heterogeneity assessment by calculating the I² statistic value (Higgins 2003). We could not perform subgroup analyses (see Subgroup analysis and investigation of heterogeneity). If there was evidence of substantial heterogeneity, defined as an I² statistic value of greater than 50%, we would investigate and report the possible reasons for this. If there was evidence of heterogeneity with an I² statistic value of greater than 75% to 80%, we would not calculate a common measurement of the effect (Deeks 2024).

Assessment of reporting biases

We did not examine funnel plots for each primary outcome to assess the potential for small study effects, because there were insufficient included studies (fewer than 10).

We attempted to obtain unpublished results, to minimise the risk of reporting bias. We did not analyse the results of the included studies by means of funnel plots to determine the possible existence of bias (fewer than 10 trials).

In addition, we have not explored possible sources of asymmetry in funnel plots (selective outcome reporting, poor methodological quality leading to spuriously inflated effects in smaller studies, true heterogeneity, artefactual and chance).

Given that the review includes fewer than 10 RCTs, for any of the primary outcomes we used a qualitative assessment of reporting biases for that outcome.

Data synthesis

If meta‐analysis was not appropriate, we stated the reason(s) why it was not appropriate. In addition, we have summarised the results of the included studies.

Sufficient, clinically similar studies were available, so we pooled their results in meta‐analyses using RevMan (RevMan 2025). We used adjusted summary statistics if available; otherwise, we used unadjusted results.

• For time‐to‐event data, we planned to pool HRs, but we did not find time‐to‐event data.

• For dichotomous outcomes, we pooled the RR for each study.

• For continuous outcomes, we planned to pool the MD values if all studies measured the outcome on the same scale. However, if studies used different measurement scales, we planned to pool SMD values. However, we did not find data on continuous outcomes.

If any trials had multiple treatment groups, we divided the ‘shared’ comparison group into the number of treatment groups and comparisons between each treatment group, and we treated the split comparison groups as independent comparisons.

We used random‐effects models with inverse variance weighting for all meta‐analyses (DerSimonian 1986).

Subgroup analysis and investigation of heterogeneity

In the protocol, we planned to perform a subgroup analysis, grouping the trials by the following.

• Histology: treatment of endometrial cancer or AEH

However, we could not perform a subgroup analysis because we only found information on AEH in the included RCTs, so we could not compare endometrial cancer and AEH.

Sensitivity analysis

We did not perform a sensitivity analysis for the primary outcome because there were insufficient studies included that took into account the risk of bias, the source of the data (type of studies included) and the decisions taken according to the heterogeneity analysis.

Summary of findings and assessment of the certainty of the evidence

We presented the overall quality of the evidence for each outcome according to the GRADE approach, which takes into account issues related to internal validity (risk of bias, inconsistency, imprecision, publication bias) and to external validity, such as directness of results (Langendam 2013). We created a summary of findings table based on the methods described in the Cochrane Handbook for Systematic Reviews of Interventions (Schünemann 2024), and using GRADEpro GDT software. We used the GRADE checklist (Meader 2014), and the GRADE Working Group certainty of evidence definitions (Hultcrantz 2017).

• High certainty: we are very confident that the true effect lies close to that of the estimate of the effect.

• Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different.

• Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect.

• Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

To interpret the findings and to rate the quality of evidence, we used the GRADE approach (Guyatt 2011). First, we analysed the overall certainty of evidence for each outcome individually, downgrading the evidence from ‘high certainty' to ‘moderate', ‘low' or ‘very low', depending on the risk of bias, indirectness of evidence, inconsistency, imprecision of effect estimates or potential publication bias. Afterwards, we took this analysis into account to draft the review conclusions. We used GRADEpro GDT software to produce summary of findings tables with the results of this analysis.

The outcomes included in the summary of findings tables are:

• overall survival;

• live birth rate;

• progression‐free survival;

• complete pathological response rate;

• severe adverse events (Grade 3+);

• quality of life; and

• surgery for persistent/progressive disease (hysterectomy).

We prioritised evidence from RCTs for presentation in the summary of findings tables, over any available NRSI evidence. The summary of findings tables show the combined evidence from women with endometrial cancer and AEH.

We included summary of findings tables for the following comparisons.

• Metformin plus progestin compared to progestin

• Levonorgestrel intrauterine system compared to oral progestin

• Oral progestin plus levonorgestrel intrauterine system compared to oral progestin

We did not produce summary of findings tables if we found no studies for a specific comparison. We plan to include summary of findings tables for the following clinically relevant comparisons in future updates if we find relevant studies.

• Hysteroscopic resection compared to oral progestin

• Weight loss plus progestin (any route) compared to progestin (any route).

Results

Description of studies

Results of the search

For details of our search strategy, see Appendix 1.

Figure 1 shows the process of screening and selecting studies. We carried out several electronic database searches between September 2018 and January 2025.

In September 2018, our electronic search identified:

• MEDLINE, 1832 references from 1946 to August, week 4 2018;

• Embase, 4024 references from 1980 to 2018 week 35; and

• CENTRAL, 2017, Issue 7, 719 references.

In September 2020, our electronic search identified:

• MEDLINE, 59 references from November 2019 to September week 3 2020;

• Embase, 80 references from November 2019 to 2020 week 39; and

• CENTRAL, 2020, Issue 9, 18 additional references.

In March 2022, our electronic search identified:

• MEDLINE, 159 references from September 2020 to March 2022;

• Embase, 429 references from September 2020 until 2022 week 9, and

• CENTRAL, 2022, Issue 3, 89 additional references.

In April 2023 our electronic search identified:

• MEDLINE, 84 references from March 2022 to 18 April 2023;

• Embase, 405 references from March 2022 to 2023, week 15);

• CENTRAL, 2023, Issue 4, 18 references.

On 3 February 2025, our electronic search identified:

• MEDLINE, 137 references from April 2023 to 31 January 2025;

• Embase, 753 references from April 2023 to 31 January 2025; and

• CENTRAL, 2024, Issue 12, 14 references.

Our electronic searches resulted in 8820 hits, with 89 from other sources. After removing 1658 duplicate records, 7251 records were left for screening. Our screening of the electronic records and studies retrieved from trials registers, conference proceedings, abstracts, co‐operative trial groups, reference lists and other sources resulted in 155 full‐text studies, which we assessed for eligibility. Of these full‐text studies, 12 met the inclusion criteria (one study was reported in three different papers; one study was reported in two references). We excluded 110 of these studies for the reasons specified in Characteristics of excluded studies. See Figure 1.

We classified 19 studies as ongoing studies (Characteristics of ongoing studies), and 11 studies as awaiting classification (Characteristics of studies awaiting classification).

Included studies

See Characteristics of included studies.

Study design

We included six RCTs (one with three articles and another with two articles) in this review (Lindahl 1991; Orbo 2014; Shan 2014; Xu 2023 A; Xu 2023 B; Yang 2020); and six NRSs studies (Acosta‐Torres 2020; Chen 2016; Mitsuhashi 2019; Park 2013; Simpson 2014; Zhou 2017).

Settings

Orbo 2014 was a multicentre study including 17 gynaecologic centres in Norway. Park 2013 was a multicentre study with eight centres. Two centres participated in each of Acosta‐Torres 2020 and Simpson 2014. Xu 2023 A, Xu 2023 B, Yang 2020, Shan 2014 and Lindahl 1991 were single‐centre RCTs. Zhou 2017 and Chen 2016 are NRSs from single centres in China. Mitsuhashi 2019 is a NRS study from a single centre.

Women were recruited at: general hospitals for Orbo 2014, Lindahl 1991 and Zhou 2017; university hospitals for Yang 2020, Shan 2014, Mitsuhashi 2019, Chen 2016, Xu 2023 A and Xu 2023 B; university oncology departments for Acosta‐Torres 2020; specialised oncology departments for Simpson 2014, and a tertiary hospital and cancer centre for Park 2013.

Participants

We identified 12 studies with 904 participants.

Condition

Randomised controlled trials

• Four RCTs included women with AEH (Lindahl 1991; Orbo 2014; Shan 2014; Xu 2023 B).

  • Lindahl 1991 included women classified as having hyperplasia, according to the Kurman and Norris 1987 classification (Kurman 1987).

  • Orbo 2014 included women with simple, complex, and atypical complex hyperplasia, using a modified WHO 1994 classification from Kurman 1985.

  • Shan 2014 did not report the classification system used, but defined the histological characteristics of AEH as in Kurman 2011.

  • Xu 2023 B classified AEH using the WHO 2014 pathological classification (Zaino 2014)

• Two RCTs studied women with adenocarcinoma stage IA endometrial cancer.

  • Xu 2023 A used the WHO 2014 classification (Zaino 2014).

  • Yang 2020 included AEH and endometrioid grade 1 endometrial cancer, using the WHO 2014 classification (Zaino 2014).

Non‐randomised studies

• One observational study included women with endometrial cancer (Park 2013).

• Five observational studies were in women with AEH and endometrial cancer (Acosta‐Torres 2020; Chen 2016; Mitsuhashi 2019; Simpson 2014; Zhou 2017).

Age

Randomised controlled trials

• Lindahl 1991 provided no information on the age or fertility desire of the women included.

• Orbo 2014 included women aged from 30 years to 70 years, with no information on their fertility desire.

• Shan 2014 included women with fertility desire with a median age of 34 years (range 26 to 42). The median age in the megestrol acetate plus metformin group was 34 years (range 33 to 42); while in the megestrol acetate group, the median age was 33 (range 26 to 42).

• Xu 2023 A included women with fertility desire with a median age of 30 years (range 21 to 43), 30 years (range 21 to 43) in the megestrol acetate group and 30 years (range 25 to 42) in the megestrol acetate plus levonorgestrel IUS group.

• Xu 2023 B included women aged 32 years (range 23 to 43) in the megestrol acetate group, 32 years (range 19 to 44) in the levonorgestrel IUS group and 33 years (range 20 to 44) in the megestrol acetate + levonorgestrel IUS group.

• Yang 2020 included women with a mean age of 32.0 years (SD ± 4.5 years) in the megestrol acetate group and 33.4 ± 5.2 (mean ± SD) in the megestrol acetate plus metformin group, and with fertility desire.

Non‐randomised studies

• Acosta‐Torres 2020 included women with a desire for fertility and a median age of 35 years (range 30 to 38.5): a median of 32 years (range 29–35) in the metformin plus progestin group, and 36 years (range 31 to 39) in the progestin group.

• Chen 2016 studied women aged (mean) 32 years (range 21 to 41) with fertility desire.

• Mitsuhashi 2019 included women with fertility desire with a median age of 35 years (range 26 to 44) in the medroxyprogesterone acetate plus metformin group, and also in the comparison group: historic cohort treated with medroxyprogesterone acetate (range 28–45).

• Park 2013 included women with a mean age of 31.3 years (range 21 to 40) and a desire for fertility.

• Simpson 2014 included women with a median age of 36.5 years (range 26 to 44), with fertility desire.

• Zhou 2017 included 29 women under 45 years of age with early, well‐differentiated endometrial carcinoma (n =17) and complex atypical hyperplasia (n = 12) and who desired fertility‐sparing treatment. The women had a mean age of 30.6 years (range 21 to 42).

Comparisons

• Four studies (two RCTs and two observational retrospective cohort studies) compared metformin plus progestin with progestin alone (Acosta‐Torres 2020; Mitsuhashi 2019; Shan 2014; Yang 2020).

• Two observational retrospective cohort studies compared the effects of two progestins: medroxyprogesterone acetate and megestrol acetate (Chen 2016; Park 2013).

• Two retrospective cohort studies compared high‐dose versus low‐dose progestins (Park 2013; Simpson 2014).

• Two RCTs compared oral progestin with levonorgestrel IUS (Orbo 2014, Xu 2023 B).

  • Orbo 2014 (three arms) compared medroxyprogesterone acetate 10 mg orally, daily, for six months; medroxyprogesterone acetate 10 mg orally, 10 days per cycle, for six months; and levonorgestrel IUS for six months. An endometrial biopsy (performed by the gynaecologist) was scheduled after 3 and 6 months.

  • Xu 2023 B compared continuous (long‐term, noncyclical) oral megestrol acetate 160 mg once daily and levonorgestrel IUS (containing 52 mg levonorgestrel) insertion, until a complete pathological response was achieved.

• Two RCTs compared oral progestin with oral progestin plus levonorgestrel IUS (Xu 2023 A; Xu 2023 B).

  • Xu 2023 A and Xu 2023 B compared continuous (long‐term, noncyclical) oral megestrol acetate (160 mg once daily) and oral megestrol acetate (160 mg once daily) plus levonorgestrel IUS (containing levonorgestrel 52 mg) insertion.

  • Xu 2023 B compared oral progestin plus levonorgestrel IUS with levonorgestrel IUS alone.

• One RCT compared the effect of progestin with endometrial curettage (Lindahl 1991).

• One observational study compared GnRH and levonorgestrel IUS with GnRH and letrozole (Zhou 2017).

Intervention mode, frequency, follow‐up and comparative conditions

Randomised controlled trials

• Lindahl 1991 treated women with medroxyprogesterone acetate 500 mg intramuscularly, twice‐weekly for three months, with abrasion performed 3, 6, 12 and 24 months after the initial diagnosis. The comparison group was treated with abrasion at 3, 9 and 24 months after the initial diagnosis. Follow‐up lasted for two to five years.

• Orbo 2014 compared medroxyprogesterone acetate 10 mg orally, 10 days per cycle, for six months to levonorgestrel IUS for six months. An endometrial biopsy (undertaken by a gynaecologist) was scheduled after three and six months.

• Shan 2014 treated the women in the intervention group with megestrol acetate (160 mg orally, daily) plus metformin (500 mg orally, three times a day) until complete pathological response. Afterwards, three months of oral contraception pills were recommended. The control group received megestrol acetate (160 mg orally, daily), followed by three months of oral contraception pills. The first biopsy control was done at 12 weeks. If a partial response was observed (hyperplasia without atypia), then the participant was treated for 12 more weeks and a new biopsy was performed. If there was no response to the treatment, women could repeat the treatment for 12 more weeks or opt for surgical treatment.

• Xu 2023 A compared continuous (long‐term, noncyclical) oral megestrol acetate (160 mg once daily) and oral megestrol acetate (160 mg once daily) plus levonorgestrel IUS (containing LNG 52 mg) insertion in early stage endometrial cancer. Hysteroscopy was scheduled to be performed every three months. According to the results of MRI and ultrasound, all suspected lesions were removed completely until no lesions were visible to the naked eye. A random endometrial biopsy was performed in the area where no obvious lesion was found. All the specimens were sent separately for pathological diagnosis. The levonorgestrel IUS was taken out during each hysteroscopic evaluation and kept from contamination. Bacilli culture was performed at the same time. The levonorgestrel IUS (or a new one) was placed in the uterine cavity after each hysteroscopic evaluation. After achieving complete pathological response, the same regimen was administered for another two to three months. For women with complete pathological response without recent desire to conceive, or those who had stopped breastfeeding after delivery, cyclic oral dydrogesterone, oral contraceptive pills, or levonorgestrel IUS were administered to prevent disease recurrence. Hysterectomy was suggested for women who had completed childbearing.

• Xu 2023 B compared continuous (long‐term, noncyclical) oral megestrol acetate (160 mg once daily) and levonorgestrel IUS (containing LNG 52 mg) or megestrol acetate (160 mg once daily) plus levonorgestrel IUS. Hysteroscopic evaluations were performed every three months until complete pathological response, by two specialists. Suspected lesions were removed completely. A random endometrial biopsy was performed in an area where no obvious lesions were found. All specimens were sent separately for pathological diagnosis. During each hysteroscopic evaluation, the levonorgestrel IUS was removed, kept from contamination, and bacilli culture was performed. The levonorgestrel IUS (or a new one) was placed in the uterine cavity after each hysteroscopic evaluation. Megestrol acetate and levonorgestrel IUS or megestrol acetate or levonorgestrel IUS were administered until complete pathological response was achieved. If women experienced unacceptable side effects, treatment was discontinued. After achieving complete pathological response, the same regimen was administered for another two to three months. For women with complete pathological response without a recent plan to conceive or those who stopped breastfeeding after delivery, cyclic oral dydrogesterone, oral contraceptive pills, or levonorgestrel IUS were administered to prevent disease recurrence. Definitive hysterectomy was suggested when women remained with stable disease after seven months of treatment, or did not achieve complete pathological response after 10 months of treatment, or had PD at any time during treatment.

• Yang 2020 treated the intervention group with megestrol acetate (160 mg orally, daily) plus metformin (500 mg orally, three times a day) until two to three months after achieving complete pathological response. The control group received megestrol acetate (160 mg orally, daily) until two to three months after achieving complete pathological response. Hysteroscopic evaluation was scheduled every three months during the therapy, with MRI every six months during the treatment (for women with endometrioid endometrial cancer). When a complete pathological response was achieved, women were followed‐up every three to six months; transvaginal ultrasound and endometrial biopsy by Pipelle were used to assess the endometrium. Follow‐up for women with endometrial cancer involved annual pelvic MRI, and blood tests to check for biomarkers (CA125 and HE4).

Non‐randomised studies

• Acosta‐Torres 2020 included different treatment regimens with progestin (megestrol acetate at 80 to 160 mg orally daily, medroxyprogesterone acetate at 10 to 40 mg orally daily, micronised progesterone 400 mg orally daily), and/or LNG‐IUD at 52 mg. Oral, intrauterine or both treatments were included (systemic 61%, oral 32% and both 8% of cases). Metformin was administered at 500 mg to 1000 mg daily. Fifty‐eight women (63%) received progestogen therapy and 34 (37%) received progestogen and metformin. Of the 34 women treated with progestogen plus metformin, 27 (80%) initiated metformin at the start of the progestin therapy, while five (15%) had been taking metformin for three months prior to progestin. The mean duration of the metformin treatment was 18.3 months (range 11 to 35.2). Treatment with metformin was discontinued during initial progestin treatment or when it ended in 25 women and continued in nine women. Women on metformin plus progestin were younger, had a higher BMI at the start of treatment, and were more likely to have diabetes mellitus and PCOS than those receiving only progestin. Furthermore, the types of progestin treatments were not equally distributed, and women in the metformin plus progestin group were more likely to have local treatment with levonorgestrel IUS than in the progestin‐only cohort. No information on follow‐up was reported in the article.

• Chen 2016 treated women with medroxyprogesterone acetate (250 to 500 mg/day orally) or megestrol acetate (160 to 480 mg/day orally) for at least six months. Once a complete pathological response was achieved, maintenance therapy was administered for three to six months. Low‐dose cyclic progestin, oral contraceptives, or levonorgestrel IUS were administered to women with no immediate childbearing plans. Response to treatment was assessed histologically using hysteroscopic biopsies or dilatation and curettage, based on the doctors’ discretion and affordability. The histological evaluation was performed every three months, except for women who had received progestin treatment for more than 12 months, in which case the interval could be extended to four to six months during the second year.

• Mitsuhashi 2019 administered medroxyprogesterone acetate (400 mg orally, daily) to the control group until complete pathological response, for a minimum of six months if no progression was observed. After six months, if complete pathological response was not observed, a further six months of medroxyprogesterone acetate treatment could be added with close follow‐up. The intervention group received medroxyprogesterone acetate (400 mg orally, daily) until complete pathological response (same protocol) and metformin (initial dose 750 mg/day orally; increased weekly by 750 mg up to 2250 mg/day if no adverse effects were observed and continued after medroxyprogesterone acetate administration). Follow‐up was scheduled every two to three months with dilatation and ­curettage biopsy (under anaesthesia) until remission. After remission, follow‐up was performed on all women every three months with pelvic examinations, tumour markers, transvaginal ultrasound, and office‐based endometrial biopsy. When a complete pathological response was observed, maintenance therapy with low‐dose oestrogen/progestin was prescribed (cyclic (21/7 days) daily administration of 1 mg norethisterone and 0.035 mg ethinylestradiol) or cyclic lower‐dose progestin (15 mg of medroxyprogesterone/day/14 days).

• Park 2013 treated women with medroxyprogesterone acetate, 30 to 1500 mg/day orally (mean: 500 mg/day), or megestrol acetate, 40 to 240 mg/day orally (mean: 160 mg/day). The mean duration of the progestin treatment for all women was eight months (range 2 to 31 months). The median length of treatment was 9.5 months. After treatment, all women were followed‐up every three to six months using pelvic examination, tumour marker or imaging studies. Usually, transvaginal ultrasound or MRI, and sometimes CT, positron emission tomography (PET) or PET–CT were used as needed. The response to progestin treatment was assessed histologically with office‐based endometrial biopsy, dilatation and curettage biopsy or hysteroscopic biopsy.

• Simpson 2014 compared treatment of 100 mg medroxyprogesterone acetate or higher or 80 mg megestrol acetate or higher daily, orally with less than 100 mg medroxyprogesterone acetate or less than 80 mg megestrol acetate daily, orally. Diagnosis and response to treatment were assessed with endometrial curettage or office‐based endometrial biopsy. Several surveillance biopsies were performed (median 4, range 2–12) at different intervals (due to the retrospective nature of this study) while on progestin therapy and after complete pathological response.

• Zhou 2017 treated women with GnRHa 3.75 mg intramuscularly every four weeks plus levonorgestrel IUS for six months. If women had no immediate fertility desire, levonorgestrel IUS was maintained. In the case of a larger uterus than eight gestational weeks, endometrium more than 20 mm deep or metrorrhagia initially, treatment was GnRHa 3.75 mg intramuscularly, every four weeks plus letrozole 2.5 mg orally, daily. After the third injection of 3.75 mg of GnRHa, the levonorgestrel IUS was inserted and left for at least six months. The women received a combination of GnRHa with letrozole if they presented with obesity, oligomenorrhea, infertility, or PCOS. Follow‐up was performed with ultrasound examination every month and endometrial sampling via hysteroscopy and curettage every three months. Women with an inserted levonorgestrel IUS also received hysteroscopy. The median follow‐up time was 18.7 months (range 5.6‐54.9 months).

Outcomes

Primary outcomes

• Overall survival: none of the studies evaluated overall survival.

• Live birth rate: six studies reported live birth rate (Acosta‐Torres 2020; Chen 2016; Mitsuhashi 2019; Xu 2023 A; Xu 2023 B; Yang 2020). We also obtained data on live birth rate from Shan 2014. Acosta‐Torres 2020 calculated the live birth rate as the number of live births for the total number of women included in the study. Yang 2020 provided data to calculate live births for the total number of women who tried to conceive, as did Shan 2014.

Secondary outcomes

• Progression‐free survival: none of the studies reported progression‐free survival. Four NRSs reported recurrence‐free survival (Acosta‐Torres 2020; Chen 2016; Mitsuhashi 2019; Park 2013). We did not evaluate this outcome.

• Complete pathological response rate: all 12 included studies reported the complete pathological response rate, using a similar definition of the outcome: complete pathological response or regression.

• Severe adverse events: three RCTs reported severe adverse events (Xu 2023 A; Xu 2023 B; Yang 2020).

• Quality of life: none of the studies reported quality of life.

• Psychological symptoms: none of the studies reported psychological symptoms.

• Pregnancy rate: four RCTs reported pregnancy rate (Shan 2014; Xu 2023 A; Xu 2023 B; Yang 2020).

• Surgery for persistent/progressive disease (hysterectomy): one RCT (Yang 2020), and one NRS (Acosta‐Torres 2020) reported this outcome. Shan 2014 provided the data to calculate the outcome surgery for persistent/progressive disease.

Study funding and declarations of interest

• Acosta‐Torres 2020 was supported by the Kosegarten Family Research Fund and the Stoltz Family Research Fund. No conflicts of interest were declared.

• Chen 2016 reported no study funding or conflicts of interest.

• Lindahl 1991 was supported in part by grants from the John and Augusta Persson Foundation for Medical Scientific Research and the University of Lund. No information on conflicts of interest is given. A secondary reference was supported in part by grants from the John and Anguta Person Foundation for Medical Scientific Research, the Medical Faculty of the University of Lund, Magnus Bergvalls Stiftelse and The Royal Physiographic Society, Lund (Lindahl 1994 A). No information on conflicts of interest is given.

• Mitsuhashi 2019 reported no conflicts of interest relevant to this article, and did not report information on study funding.

• Orbo 2014 received funding from the Norwegian Cancer Association, the Regional Research Board of Northern Norway (Helse Nord), the Bank of North Norway, and the University of Tromsø. This study also received fees from Bayer for invited lectures given at scientific seminars for gynaecologists (Orbo 2014). For other contributing authors, ABV, megestrol acetate, IP and BS, no disclosure of interest exists.

• Park 2013 no financial support was received, and there were no conflicts of interest.

• Shan 2014 was supported by the National Natural Science Foundation of China, 2012 (NSFC No.: 81101953), Shanghai Municipal Science Foundation, 2013 (Project No.: 134119a4500), National Natural Science Foundation of China (NSFC No.: 81210108021), and Shanghai Municipal Science Foundation, 2011 (Project No.: 11ZR1404300). No potential conflict of interest relevant to this article was reported.

• Simpson 2014 received financial support from The Mount Sinai Hospital/University Health Network Research Grant. No conflicts of interest are reported.

• Xu 2023 A was supported by the National Key Research and Development Program of China (Grant No. 2019YFC1005200 and 2019YFC1005204), Shanghai Medical Centre of Key Programs for Female Reproductive Diseases (Grant No. 2017ZZ010616) and Shen Kang clinical project (SHDC22021219).

• Xu 2023 B was supported by the National Key Research and Development Program of China (Grant No. 2019YFC1005200 and 2019YFC1005204), Shanghai Medical Centre of Key Programs for Female Reproductive Diseases (Grant No. 2017ZZ010616), Shanghai sailing program (Grant No. 19YF1404200), and Shen Kang clinical project (SHDC22021219).

• Yang 2020 was funded by the National Key Technology R&D Programme of China (Grant Nos. 2019YFC1005200 and 2019YFC1005204), National Natural Science Foundation of China (grant nos 81671417 and 81370688), Shanghai Medical Centre of Key Programmes for Female Reproductive Diseases (grant no. 2017ZZ010616), Shanghai Science and Technology Development medical guide project (grant nos 17411961000, 134119a4500, 19411960400) and Municipal Human Resources Development Programme for Outstanding Leaders in Medical Disciplines in Shanghai (grant no. 2017BR035). Authors declared no conflicts of interest.

• Zhou 2017 did not report information on study funding or conflicts of interest.

Excluded studies

From the 155 studies assessed for full‐text eligibility, we excluded 110 for the following reasons:

• 58 studies used an ineligible study design.

• 30 studies included an ineligible patient population.

• 7 studies were duplicated.

• 7 studies included ineligible outcomes.

• 3 studies included an ineligible comparator.

• 2 studies were carried out in an ineligible setting.

• 1 study was not available as a full text: we wrote to the authors but received no response.

• 1 study was a corrigendum of a list of authors.

• 1 study investigated an ineligible intervention.

See Characteristics of excluded studies.

We found 11 references corresponding to studies awaiting classification (see Characteristics of studies awaiting classification), and 19 references corresponding to ongoing studies (see Characteristics of ongoing studies).

Risk of bias in included studies

Randomised controlled trials

We assessed risk of bias only for the six included RCTs. Our risk of bias judgements are shown in Figure 2 and Figure 3.

• We classified five studies as high risk for overall risk of bias:

  • Lindahl 1991: we judged one domain as high risk and three as unclear risk of bias;

  • Orbo 2014: we judged one domain as high risk and the remaining as low risk of bias;

  • Shan 2014: we judged two domains as unclear risk and two as high risk of bias;

  • Xu 2023 A and Xu 2023 B: we judged one domain as high risk, and one as unclear risk;

• and one study as unclear risk for overall risk of bias:

  • Yang 2020: we judged one domain as unclear risk of bias.

Allocation

Sequence generation

Four studies specified the generation of random sequences. Yang 2020 and Orbo 2014 generated the sequence with a computer‐based procedure of simple randomisation. Orbo 2014 used a computer random number generator with two strata and fixed block size. Xu 2023 A and Xu 2023 B used random number tables. We classified all these studies as being at low risk of bias.

The method for sequence generation was unclear for the remaining studies (Lindahl 1991; Shan 2014).

Allocation to conditions

One study allocated conditions with a central telephone randomisation unit to secure concealed allocation (Orbo 2014). Yang 2020 specified that treatment assignment was concealed. We classified both studies as being at low risk of bias. Two studies did not report allocation concealment, so we classified them as having an unclear risk of bias related to allocation (Lindahl 1991; Shan 2014). In Xu 2023 A, the treatment allocation was concealed before the women were enroled. In Xu 2023 B, the treatment allocation was concealed before the women were successfully enroled. The studies did not specify the method used to conceal allocation; we classified both studies as unclear risk of bias.

Blinding

Blinding of participants and personnel

Blinding of participants and personnel was difficult, given the nature of the intervention and the different doses or treatment routes (levonorgestrel IUS, oral progestin, etc). We considered one study as being at low risk of bias because the hysteroscopists who performed the exams and biopsies were not aware of the assignment of patients (Yang 2020). However, patients and physicians in the study were probably aware of allocation.

Xu 2023 A and Xu 2023 B are open‐label RCTs and all study physicians and patients were aware of the treatment assignment. None of the clinicians who performed the hysteroscopic evaluation of the women in these trials were aware of the treatment allocations. There was incomplete blinding, but we judged that the outcome is not likely to be influenced by lack of blinding. We classified both studies as low risk of bias.

There was no information on the blinding of participants in the remaining studies, but given the different routes of administration or dose frequency, we considered that there was probably no effective blinding (Lindahl 1991; Orbo 2014; Shan 2014). However, we judged that the outcome is not likely to be influenced by lack of blinding. We classified all three studies as low risk of bias.

Blinding of outcome assessors

Four studies clearly reported that outcome assessors were blinded (Orbo 2014; Xu 2023 A; Xu 2023 B; Yang 2020). Another study did not report if outcome assessors were blinded but provided information on the assessment process: “Pathological diagnosis was confirmed by two experienced gynaecologic pathologists. If their opinion differed, a seminar was held in the Department for the final diagnosis” (Shan 2014). This process was performed to avoid detection bias. We classified all these studies as being at low risk of bias.

The remaining study did not provide information on the blinding of outcome assessors (Lindahl 1991). We judged this study at unclear risk of bias for this domain.

Incomplete outcome data

Xu 2023 A at 16 weeks, complete pathological response, had 14.2% of missing cases. Missing cases were relatively high. However, missing outcome data were balanced in numbers across intervention groups, with similar reasons for missing data across groups. Xu 2023 B, for complete pathological response rate, reported a relatively high missing rate (26.7% and 15.6% at 16 and 32 weeks of treatment, respectively). However, missing outcome data were balanced in numbers across intervention groups, with similar reasons for missing data across groups.

Two studies were at a high risk of attrition bias because of a high dropout rate of more than 15% (Shan 2014; Yang 2020). One study reported 27% of missing cases (Shan 2014). Therefore, we classified Shan 2014 as being at high risk of incomplete outcome data. Yang 2020 reported that, in the outcome complete pathological response at 16 weeks, there were 20% missing cases or exclusions in the control (progestin) group, and 11.9% in the experimental (metformin plus progestin) group. The reasons for missing cases were similar in both groups: not attending hysteroscopy or surgery. At 32 weeks, missing cases were 7/73 (9.5%) in the progestin group and 5/75 (6.6%) in the metformin plus progestin group. At 12 months, missing cases were: 6/74 and 3/72 (8.1 % and 4.1% respectively). The 'high risk' judgement is applicable for the outcome of complete pathological response at 16 weeks but is not applicable to the other outcomes. For this reason, our judgement for this domain is unclear risk of bias.

The two remaining studies were at low risk of bias for this domain (Lindahl 1991; Orbo 2014).

Selective reporting

Six studies had a low risk of selective reporting (Lindahl 1991; Orbo 2014; Shan 2014; Yang 2020; Xu 2023 A; Xu 2023 B). They reported all prespecified outcomes in the results section of their paper. Yang 2020, Xu 2023 A and Xu 2023 B reported that a protocol of the trial had been published previously (see secondary references to the included studies). Changes in the protocol were specified, and probably had no impact on the results of the studies.

Other potential sources of bias

We identified five studies as having other potential sources of bias. In five studies, there was a small sample size. Shan 2014 was a randomised pilot study and included few women. Lindahl 1991 and Orbo 2014 were studies that included all types of endometrial hyperplasia and had few cases of AEH. Xu 2023 A and Xu 2023 B had a small sample size that did not reach the calculated number.

• Lindahl 1991 included very little information on patient characteristics, and did not report the baseline imbalance of the groups.

• Orbo 2014 included 6% of postmenopausal women and did not specify women's desire regarding fertility.

• Shan 2014 was performed in a selected subpopulation (women with AEH and at least one metabolic syndrome criterion), which constitutes a potential source of bias.

• Xu 2023 A had a small sample size that did not reach the calculated number; therefore, the power of the test was insufficient. The follow‐up time was short. The treatment groups were combined with hysteroscopic evaluation and resection of endometrial lesions, which might have concealed the difference in efficacy of the regimens.

• Xu 2023 B combined all three treatment groups with hysteroscopic evaluation and resection of endometrial lesions, which might have concealed the difference in efficacy of the regimens. The study had a potential source of bias related to the specific study design used. The sample size did not reach the calculated number; therefore, the power of the test was insufficient. The resection of endometrial lesions might have concealed the difference in efficacy of the regimens.

Effects of interventions

See: Table 1; Table 2; Table 3

In this section we report the evidence from meta‐analysis of two included RCTs (Shan 2014; Yang 2020). See Analysis 1.1; Analysis 1.2; Analysis 1.3; Analysis 1.4.

1.1. Analysis.

Comparison 1: Metformin plus progestin compared to progestin, Outcome 1: Metformin plus progestin compared to progestin. Live birth rate. RCTs

1.2. Analysis.

Comparison 1: Metformin plus progestin compared to progestin, Outcome 2: Metformin plus progestin compared to progestin. Complete pathological response rate (12‐16 weeks). RCTs

1.3. Analysis.

Comparison 1: Metformin plus progestin compared to progestin, Outcome 3: Metformin plus progestin compared to progestin. Pregnancy rate. RCTs

1.4. Analysis.

Comparison 1: Metformin plus progestin compared to progestin, Outcome 4: Metformin plus progestin compared to progestin. Surgery for persistent/progressive disease (hysterectomy). RCTs

We also provide a narrative description of the four RCTs not combined in a meta‐analysis (Lindahl 1991; Orbo 2014; Xu 2023 A; Xu 2023 B), and the six NRSs found in this systematic review (Acosta‐Torres 2020; Mitsuhashi 2019, Chen 2016; Park 2013; Simpson 2014; Zhou 2017). A meta‐analysis of all included studies was not possible because of the heterogeneity of the interventions, outcome measures, treatment regimens, outcome time points, population baseline characteristics and variation in methodological certainty. We have included a summary of the evidence without meta‐analysis with the results of these studies for the outcomes covered by the scope of this review (Table 4).

1. Summary of the comparisons without meta‐analysis.

Study	Outcome	Type of study	Type of lesion	Intervention (n(N)	Comparison (n/N)	RR (95% CI)	Risk of bias	Comments
Metformin (MTF) plus progestin (PRG) compared to progestin
Acosta‐Torres 2020 Mitsuhashi 2019	Complete pathological response rate	NRSI NRSI	AEH+EC AEH+EC	23/34 61/63	40/58 20/23	0.98 (0.73‐1.31) 1.11 (0.94‐1.31)	Moderatea Seriousa
Acosta‐Torres 2020	Live birth rate		AEH+EC	2/34	14/58	0.2437 (0.0589 ‐ 1.0080)	Moderatea	P = 0.04 (Chi2 ) Data reported on the total number of participants
Medroxyprogesterone acetate (MPA) compared to megestrol acetate (MA)
Chen 2016 Park 2013		NRSI NRSI	AEH+EC EC	MPA	MA
Chen 2016 Park 2013	Live birth rate			‐ ‐	‐ ‐		Seriousa Seriousa	Data reported globally
Chen 2016 Park 2013	Complete pathological response rate			23/32 70/91	17/21 45/57	0.89 (0.66‐1.20) 0.97 (0.82‐1.16)	Seriousa Seriousa
Progestin high dose compared to progestin low dose
				High dose	Low dose
Park 2013 Simpson 2014	Complete pathological response rate	NRSI NRSI	EC AEH+EC	59/75 ‐	56/73 ‐	1.03 (0.86‐1.22)	Seriousa Seriousa	Primary outcomes not evaluated HR for high‐dose PRG = 0.92 (95% CI 0.36 to 2.33)
Levonorgestrel intrauterine system (LNG‐IUS) compared to oral progestin
Xu 2023 B Orbo 2014		RCT RCT	AEH AEH	LNG‐IUS	Oral progestin
Xu 2023 B	Live birth rate			8/16	5/18	1.80 (0.74‐4.39)	Highb	Not evaluated by Orbo 2014
Xu 2023 B Orbo 2014	Complete pathological response rate			21/46 6/6	11/43 10/13	1.78 (0.98‐3.25) 1.30 (0.97‐1.75)	Highb Highb	CR at 16 weeks, age 32 years (23–43), single‐centre, university hospital Age 30‐70 years, 6% postmenopausal, multicentre, gynaecological centres
Xu 2023 B	Severe adverse events grade 3 weight gain			2/60	10/58	0.19 (0.04‐0.84)	Highb	Orbo 2014 reported data globally for all types of hyperplasia. No grade 3 adverse events were reported
Xu 2023 B	Pregnancy rate			13/16	12/ 18	1.22 (0.81‐1.82)	Highb
	Surgery for persistent/ progressive disease							Data reported globally
Gonadotropin‐releasing hormone agonist (GnRHa) combined with levonorgestrel intrauterine system (LNG‐IUS) or letrozole
Zhou 2017	Complete pathological response rate	NRSI	AEH+EC				Seriousa	HR = 0.957(95% CI = 0.850 to 1.077)
Endometrial curettage with or without oral progestin
Lindahl 1991		RCT	AEH	Endometrial curettage with oral progestin	Endometrial curettage
	Progression‐free survival	Not evaluated						Progression with MPA + curettage (11/11) vs curettage alone (14/15) RR 1.07 (0.94‐1.23) after 3 months
	Complete pathological response rate			9/10	6/15	2.25 (1.17‐4.32)	Highb
	Surgery for persistent/ progressive disease			3/9	3/15	1.67 (0.42‐6.56)	Highb
Oral progestin plus levonorgestrel intrauterine system (LNG‐IUS) compared to oral progestin
Xu 2023 A Xu 2023 B		RCT RCT	EC AEH	Oral progestin plus LNG‐IUS	Oral progestin			(A) Mean age 30 years (21‐43), university hospital, follow‐up: 31.6 months (B) Mean age 33 years (19–44), university hospital, follow‐up: 27.8 months
Xu 2023 A Xu 2023 B	Live birth rate	RCT RCT	EC AEH	4/12 5 /13	5/21 5/18	1.40 (0.46‐4.24) 1.38 (0.50‐3.82)	Highb Highb
Xu 2023 A Xu 2023 B	Complete pathological response rate			7/28 16/43	5/26 11/43	1.30 (0.47‐3.59) 1.45 (0.77‐2.76)	Highb Highb	At 16 weeks At 16 weeks, hysteroscopic resection was performed
Xu 2023 A Xu 2023 B	Severe adverse events grade 3 weight gain			2/28 9/54	2/31 10 /58	1.11 (0.17‐7.34) 0.97 (0.43‐2.20)	Highb Highb	Weight gain was the only grade 3 side effect
Xu 2023 A Xu 2023 B	Pregnancy rate			9/12 11/13	13/21 12/18	1.21 (0.76‐1.94) 1.27 (0.85‐1.89)	Highb Highb
Xu 2023 A Xu 2023 B	Surgery for persistent/ progressive disease			2/29 1/43	1/30 1/43	2.07 (0.20‐21.60) 1.00 (0.06‐15.48)	Highb Highb	No progression was observed in either study
Oral progestin plus levonorgestrel intrauterine system (LNG‐IUS) compared to LNG‐IUS
Xu 2023 B		RCT	AEH	Oral progestin plus LNG‐IUS	LNG‐IUS
	Live birth rate			5/13	8/16	0.77 (0.3 3‐1.79)	Highb
	Complete pathological response rate			16/43	21/46	0.82 (0.49‐1.34)	Highb	All the participants underwent hysteroscopic resection of the lesions.
	Severe adverse events: weight gain			9/54	2/60	5.0 (1.13‐22.13)	Highb
	Pregnancy rate			11/13	9/54	5.08 (2.68‐9.63)	Highb
	Surgery for persistent/ progressive disease			1/55	0/58	Not estimable	Highb	31 women with SD or PR refused surgery and followed the same treatments or alternative until CR
AEH: atypical endometrial hyperplasia; CR: complete response; EC: endometrial cancer; MA: megestrol acetate; MPA: medroxyprogesterone acetate; MTF: metformin; n: cases with events; N: number of participants; NRSI: non‐randomised studies of interventions; PR: partial response; PRG: progestin; RCT: randomised controlled trial; SD: stable disease

^aWe assessed risk of bias in non‐randomised studies using ROBINS I (Sterne 2016). 
^bWe assessed risk of bias in randomised controlled trials using the Cochrane risk of bias tool, RoB 1 (Higgins 2017).

Of the 12 studies included, we considered 10 (one study in two comparisons) to be sufficiently homogeneous to combine them in six different comparisons:

• Yang 2020 and Shan 2014 are RCTs that compared metformin plus progestin with progestin;

• Acosta‐Torres 2020 and Mitsuhashi 2019 are retrospective NRSs that compared metformin plus progestin with progestin;

• Chen 2016 and Park 2013 (NRSs) compared medroxyprogesterone acetate with megestrol acetate;

• Simpson 2014 and Park 2013 (NRSs) compared high‐dose versus low‐dose progestin.

We narratively describe the effects of interventions of these NRSs.

Two RCTs compared oral progestin to levonorgestrel IUS (Xu 2023 B; Orbo 2014), which we did not combine for meta‐analysis because of the difference in the characteristics of both studies: Xu 2023 B treated all women with hysteroscopic resection, apart from the treatments studied. Orbo 2014 included all types of endometrial hyperplasia and had few cases of AEH; in addition, it included 6% of postmenopausal women.

Two RCTs compared oral progestin with oral progestin plus levonorgestrel IUS (Xu 2023 A; Xu 2023 B). We summarised the results of this comparison narratively. We did not combine them for meta‐analysis because of the clinical diversity of the studies.

Xu 2023 B (RCT) compared levonorgestrel IUS with oral progestin plus levonorgestrel IUS. We summarised the results of this study narratively.

Finally, in the following comparisons we found a few studies that are described narratively:

• GnRHa combined with levonorgestrel IUS or letrozole (NRS; Zhou 2017);

• endometrial curettage with or without oral progestin (RCT; Lindahl 1991).

The comparisons hysteroscopic resection compared to oral progestin, and weight loss plus progestin (any route) compared to progestin (any route) are clinically relevant. In these comparisons, no study is included.

1. Metformin plus progestin compared to progestin

See Table 1

Primary outcomes

1.1. Overall survival

This outcome was not evaluated in any study.

1.2. Live birth rate

See Analysis 1.1; Figure 4.

4.

Forest plot of comparison: 1 Metformin plus progesstin compared to progestin. Atypical endometrial hyperplasia and endometrial cancer, outcome: 1.2 Live birth rate

Two RCTs (Shan 2014; Yang 2020), and two NRS (Acosta‐Torres 2020; Mitsuhashi 2019), provided information on live births. It was possible to combine the results from the RCTs with unpublished data after we contacted the authors to request additional information.

1.2.1. Randomised controlled trials

We pooled data from the two RCTs that calculated the percentages of live births from the number of women who tried to conceive (Shan 2014; Yang 2020; 72 women). In our pooled analysis, metformin plus progestin may result in little to no difference in live birth rate after treatment of endometrial cancer and AEH (inverse variance (IV), random‐effects (RE); RR 1.80, 95% CI 0.88 to 3.68, P = 0.11, I² = 0%; 2 studies, 72 women; low‐certainty evidence; Analysis 1.1). The absolute effects of the intervention were 180 more live births per 1000 (95% CI, from 27 fewer to 603 more). The overall certainty of evidence is low, downgraded two levels due to imprecision of the results (Table 1).

1.2.2. Non‐randomised studies

Acosta‐Torres 2020 found, in the metformin plus progestin group, 2/24 live births (6%), and 14/58 (24%) in the progestin group (RR 0.243, 95% CI 0.0589 to 1.0080; P = 0.04; Table 4). The study authors compared the events with the total women in each group, not with women who tried to get pregnant.

Mitsuhashi 2019 only reported data on live births for the metformin plus medroxyprogesterone acetate group. Therefore, we could not make any comparisons between these two arms of treatment regarding this outcome (Table 4).

Secondary outcomes

1.3. Progression‐free survival

This outcome was not evaluated in any study.

1.4. Complete pathological response rate

1.4.1. Randomised controlled trials

See Analysis 1.2; Figure 5.

5.

Forest plot of comparison: 1 Metformin plus progestin compared to progestin. Atypical endometrial hyperplasia and endometrial cancer, outcome: 1.1 Complete pathological response rate

Two RCTs that evaluated metformin plus progestin versus progestin analysed the complete pathological response rate (Shan 2014; Yang 2020). The two studies included 141 women. We decided to combine the outcome measured at 16 weeks for meta‐analysis because the timing of this measurement was more homogeneous across the groups.

The evidence from the RCTs suggests that the addition of metformin to progestin may result in a slight increase in the complete pathological response (IV, RE; RR 1.85, 95% CI 1.07 to 3.19; P = 0.03, I² = 0%; 2 studies, 141 women; low‐certainty evidence; Analysis 1.2). The absolute effect is 180 more complete pathological responses per 1000 cases (95% CI from 15 more to 465 more complete responses). The overall certainty of the evidence is low, downgraded two levels due to imprecision (Table 1).

1.4.2. Non‐randomised studies

Acosta‐Torres 2020 described a median time to complete pathological response at 4.9 months. Complete pathological response was obtained in 23/34 (67.6%) women in the metformin plus progestin group and 40/58 (68.9%) in the progestin group (RR 0.98, 95% CI 0.73 to 1.31; Table 4).

Mitsuhashi 2019 observed a median complete pathological response at six months. After treatment, 61/63 (96.8%) women reached a complete pathological response to metformin plus progestin whereas it was reached in 20/23 (86.9%) women in the progestin‐only arm (RR 1.11, 95% CI 0.94 to 1.31; Table 4).

1.5. Severe adverse events

We found limited severe adverse events (grade 3+). We narratively report the results of the evidence of this outcome (adverse events grade 1 and 2) for the two studies that provided data on adverse events (Yang 2020 (RCT); Mitsuhashi 2019 (NRS)). The other two studies did not evaluate adverse events (Acosta‐Torres 2020; Shan 2014). Therefore, it was not possible to combine the data in a meta‐analysis of this outcome.

Weight gain

Yang 2020 (RCT): the most frequent adverse effect was weight gain (grade 1‐4). It was observed in 29 out of 76 women in the metformin plus megestrol acetate group. The mean weight increase was 2.5 kg (range ‐1 to 6). In the megestrol acetate group, 34 out of 74 women experienced weight gain, with a mean gain of 5 kg (range 0 to 10; P = 0.01). Differences were observed in the amount of weight increase, but not in the frequency of the outcome. See Table 1. Yang 2020 reported 5/74 (6.8%) cases of grade 3‐4 weight gain in the megestrol acetate‐treated group versus 2/76 (2.6%) in the metformin plus megestrol acetate group, (RR 0.39; 95% CI 0.08 to 1.95; 1 RCT; 150 women; low ‐certainty evidence; Analysis 1.5).

1.5. Analysis.

Comparison 1: Metformin plus progestin compared to progestin, Outcome 5: Metformin plus progestin compared to progestin. Severe adverse events grade 3‐4.

The study also reported grade 3‐4 adverse events: leukocytosis in 2/74 (2.7%) in the megestrol acetate group and 1/76 (1.3%) in the metformin plus megestrol acetate group; hypercoagulable state in 1/74 (1.4%) in the megestrol acetate group and 0/76 in the metformin plus megestrol acetate group; facial oedema in 1/76 (1.3%) in the metformin + megestrol acetate group and 0/74 in the megestrol acetate group.

Mitsuhashi 2019 (NRS): no weight gain was observed. Data were reported only in the metformin plus medroxyprogesterone acetate group.

Other adverse effects

Yang 2020 (RCT): diarrhoea grade 1‐2 was observed in 12 out of 76 women (15.8%) treated with metformin plus megestrol acetate. In the megestrol acetate group, 3/74 (4.1%) cases were reported. So, in the group metformin plus megestrol acetate more cases of diarrhoea (grade 1‐2) were reported than in the megestrol acetate group (P = 0.003). Other adverse events were uterine bleeding (7.9% versus 17.6%), nocturia (0% versus 4.1%) and mastodynia (4.0% versus 10.8%) in the metformin plus megestrol acetate and megestrol acetate groups, respectively.

Mitsuhashi 2019 (NRS) only reported adverse events in the metformin plus megestrol acetate group: six out of 61 women had nausea grade 1‐2; six out of 61 women had diarrhoea grade 1; while in three out of 61 women diarrhoea grade 2 was found. When the dose of metformin was reduced from 2250 mg to 1500 mg/day, the symptoms of diarrhoea and nausea disappeared. No weight gain was observed.

1.6. Quality of life

This outcome was not evaluated in any study.

1.7. Psychological symptoms

This outcome was not evaluated in any study.

1.8. Pregnancy rate

Two RCTs (Shan 2014; Yang 2020), and two NRSs (Acosta‐Torres 2020; Mitsuhashi 2019), provided information on pregnancy rates. We included unpublished data from Shan 2014 after we contacted the study authors for additional information.

1.8.1. Randomised controlled trials

See Analysis 1.3; Figure 6.

6.

Forest plot of comparison: 1 Metformin plus progestin compared to progestin. Atypical endometrial hyperplasia and endometrial cancer, outcome: 1.3 Pregnancy rate

The study carried out by Yang 2020 and the data from Shan 2014 calculated the percentages from the number of women who tried to conceive.

Metformin plus progestin may result in little increase to no difference in pregnancy rate (IV/RE; RR 1.05, 95% CI 0.65 to 1.68; P = 0.85, I² = 0%; 2 RCTs, 72 women; low‐certainty evidence). The absolute effect of the intervention was 24 more pregnancies per 1000 women (95% CI from 165 fewer to 320 more). The overall certainty of the evidence is low, downgraded two levels due to imprecision of the results.

1.8.2. Non‐randomised studies

Acosta‐Torres 2020 provided data on pregnancies aggregated over the two intervention groups and Mitsuhashi 2019 only reported data on pregnancies for the group metformin plus medroxyprogesterone acetate. Therefore, no comparisons could be performed on the different treatment arms.

1.9. Surgery for persistent/progressive disease (hysterectomy)

Two RCTs (Shan 2014; Yang 2020), and two NRSs (Acosta‐Torres 2020; Mitsuhashi 2019), provided information on the need for surgical treatment.

1.9.1. Randomised controlled trials

See Analysis 1.4; Figure 7.

7.

Forest plot of comparison: 1 Metformin plus progestin compared to progestin. Atypical endometrial hyperplasia and endometrial cancer, outcome: 1.4 Surgery for persistent/progressive disease (hysterectomy)

Shan 2014 reported three hysterectomies, but did not report in which treatment group they occurred. After contacting the study authors, we obtained the following data: two surgical interventions were performed in the megestrol acetate group, and one in the metformin plus megestrol acetate group.

Metformin plus progestin may result in little decrease to no difference in the need for surgery for persistent/progressive disease in endometrial cancer and AEH (IV/RE; RR 0.96, 95% CI 0.24 to 3.78; P = 0.95, I² = 0%; 2 RCTs, 166 women; low‐certainty evidence). The absolute effect was two fewer women who needed surgery per 1000 (95% CI from 37 fewer to 136 more). The overall certainty of the evidence was low, downgraded two levels due to imprecision (Table 1).

1.9.2. Non‐randomised studies

Acosta‐Torres 2020 provided data on surgery for persistent/progressive disease aggregated over the two intervention groups, and Mitsuhashi 2019 only provided data on surgery for persistent/progressive disease in the metformin plus medroxyprogesterone acetate group. It was not possible to compare treatment groups.

Subgroup analysis

In our protocol, we planned to do a subgroup analysis by grouping the studies according to histology: endometrial cancer or AEH. We could not carry out a subgroup analysis in the comparison metformin plus progestin compared to progestin (complete pathological response rate). In RCTs, we only had information on AEH, and not on endometrial cancer.

2. Levonorgestrel intrauterine system compared to oral progestin

Two RCTs provided data for the comparison of oral progestin to levonorgestrel IUS (Orbo 2014; Xu 2023 B). Both studies evaluated women with AEH. However, they also included hysteroscopic resection of the lesions at the beginning of the study and during follow‐up. In Xu 2023 B all women underwent co‐intervention (hysteroscopic resection of the lesions). Since co‐interventions are balanced in all study arms, we consider that there is no indirectness issue. Orbo 2014 analysed all types of hyperplasia and had few cases of AEH (19 cases). In Xu 2023 B the women were younger and had fertility desire (median age 33 years (range 19–44)). In Orbo 2014 only 36.5% of the women included were younger than 45 years, and 26.3% were older than 52 years; they also included postmenopausal women (6%).

Due to the clinical and methodological diversity of these studies (population and interventions), we decided not to pool these results for meta‐analysis (McKenzie 2023).

Primary outcomes

2.1. Overall survival

This outcome was not evaluated in any study.

2.2. Live birth rate

Xu 2023 B is a phase II RCT, including 180 women with AEH. Of these, 120 women were treated with megestrol acetate or levonorgestrel IUS (60 in each group); 47 women planned for parenthood. After achieving complete pathological response, in the levonorgestrel IUS group, 8/16 women (50%) had a live birth. In the megestrol acetate group, 5/18 women (27.8%) had a live birth. These results suggest that levonorgestrel IUS may result in little to no difference in the live birth rate compared to oral progestin (RR 1.80, 95% CI 0.74 to 4.39; 1 study, 34 women; low‐certainty evidence; Analysis 2.1). The absolute effects are 222 more live births per 1000 (95% CI from 72 fewer women to 942 more per 1000). The overall certainty of the evidence is low, downgraded two levels due to imprecision (Table 4; Table 2).

2.1. Analysis.

Comparison 2: Levonorgestrel intrauterine system (LNG‐IUS) compared to oral progestin, Outcome 1: Levonorgestrel intrauterine system compared to oral progestin. Live birth rate

Orbo 2014 did not report any data for this outcome.

Secondary outcomes

2.3. Progression‐free survival

This outcome was not evaluated in any study.

2.4. Complete pathological response rate

Given the clinical and methodological diversity of the studies, we decided not to pool the data for meta‐analysis. Thus, we describe the results narratively, and we present data in Table 4 (Orbo 2014; Xu 2023 B).

Orbo 2014, an RCT with 170 women with endometrial hyperplasia and three treatment arms (oral cyclic medroxyprogesterone acetate, oral continuous (long‐term, noncyclical) medroxyprogesterone acetate and levonorgestrel IUS) included a subgroup of AEH with 19 women: 6/6 women in the levonorgestrel IUS and 10/13 in the oral progestin group presented complete pathological response. The evidence suggests that levonorgestrel IUS may result in little to no difference in the complete pathological response rate compared to oral progestin (RR 1.24, 95% CI 0.86 to 1.78; 1 RCT, 19 women; low‐certainty evidence; Analysis 2.4). The absolute effects were 231 more complete pathological response cases per 1000 (95% CI from 23 fewer to 577 more complete pathological responses). The overall certainty of the evidence is low, downgraded two levels due to imprecision (Table 4; Table 2).

2.4. Analysis.

Comparison 2: Levonorgestrel intrauterine system (LNG‐IUS) compared to oral progestin, Outcome 4: Levonorgestrel intrauterine system compared to oral progestin. Complete pathological response rate.

Xu 2023 B, a phase II RCT, with 180 women with AEH and three treatment arms (60 women per arm), studied complete pathological response rate at 16 and 32 weeks. The 16‐week complete pathological response rates were 25.6% (11/43) in the megestrol acetate group, and 45.7% (21/46) in the levonorgestrel IUS group (RR 1.78, 95% CI 0.98 to 3.25; 1 RCT, 89 women; low‐certainty evidence; Analysis 2.2). Results suggest that levonorgestrel IUS may result in little to no difference in complete pathological response compared to oral progestin. The absolute effects are 200 more complete pathological response cases per 1000 (95% CI from 5 fewer to 576 more complete pathological response cases. The overall certainty of the evidence is low, downgraded two levels due to imprecision (Table 2; Table 4).

2.2. Analysis.

Comparison 2: Levonorgestrel intrauterine system (LNG‐IUS) compared to oral progestin, Outcome 2: Levonorgestrel intrauterine system compared to oral progestin.Complete pathological response rate.

The 32‐week complete pathological response rates were 78.3% (36/46) and 82.7% (43/52) in the megestrol acetate and levonorgestrel IUS groups, respectively.

2.5. Severe adverse events

Orbo 2014 reported data by treatment group but included all types of hyperplasia. No grade 3+ adverse events were observed. In the group receiving continuous (long‐term, noncyclical) 10 mg medroxyprogesterone acetate orally, 28% of women had grade 1 and 33% grade 2 irregular bleeding; 32% had grade 1 and 9% grade 2 pain; 25% had grade 1 and 37.5% (3 cases) grade 2 nausea. In the cyclic 10 mg medroxyprogesterone acetate orally group: 37.3% reported grade 1 and 20% grade 2 irregular bleeding; 36.1% grade 1 and 63.6% (7 cases) grade 2 pain; 43.7% grade 1 and 62.5% (5 cases) grade 2 nausea. In the levonorgestrel IUS group, the authors observed 34.3% grade 1 and 45.3% grade 2 irregular bleeding; 31% grade 1 and 27% grade 2 pain; 31.3% grade 1 and no grade 2 nausea.

Xu 2023 B (180 women, 60 women per arm) observed no treatment‐related deaths or serious adverse events (grade 4) during the study. Considering grade 3 weight gain, women in the levonorgestrel IUS group achieved less weight gain (2/60 women) compared to the megestrol acetate group (10/58 women). Based on the evidence from Xu 2023 B, levonorgestrel IUS may slightly reduce grade 3 weight gain compared to oral progestin (RR 0.19, 95% CI 0.04 to 0.84; 1 RCT, 118 women; low‐certainty evidence; Analysis 2.3). The absolute effects were 140 fewer weight gain cases per 1000 (95% CI from 166 fewer to 28 fewer weight gain cases). The overall certainty of the evidence is low, downgraded two levels due to imprecision (Table 2; Table 4).

2.3. Analysis.

Comparison 2: Levonorgestrel intrauterine system (LNG‐IUS) compared to oral progestin, Outcome 3: Levonorgestrel intrauterine system compared to oral progestin.Severe adverse events (weight gain)

The levonorgestrel IUS group achieved less weight gain (median, 0.0 kg; interquartile range [IQR], 0.0– 10.0, P < 0.001) compared with the megestrol acetate group (median, 5.0 kg; IQR, 0.0–0.0).

Fewer women in the levonorgestrel IUS group experienced increased nocturnal urine grade 1‐2, 13 (21.7%) versus 29 (50.0%) (P = 0.002), night sweats grade 1‐2, 9 (15.0%) versus 23 (39.7%) (P = 0.004), insomnia grade 1‐2, 8 (13.3%) versus 21 (36.2%) (P = 0.005) or facial oedema grade 1‐2, 3 (5.0%) versus 14 (24.1%) (P = 0.004) compared with the megestrol acetate group. Vaginal haemorrhage grade 1‐2 was 11 (19.0%) and 16 cases (26.7%) (P = 0.383) in the megestrol acetate and levonorgestrel IUS groups.

Xu 2023 B classified and graded adverse events using the National Cancer Institute Common Toxicity Criteria version 4.0. Orbo 2014 did not provide any information.

2.6. Quality of life

This outcome was not evaluated in any study.

2.7. Psychological symptoms

This outcome was not evaluated in any study.

2.8. Pregnancy rate

Of the 89 women in Xu 2023 B with AEH who achieved complete pathological response, 47 planned for parenthood. The pregnancy rate was 76.6% (36/47) in total, 66.7% (12/18) in the megestrol acetate group and 81.3% (13/16) in the levonorgestrel IUS group (RR 1.22, 95% CI 0.81 to 1.82; 1 RCT, 34 women; low‐certainty evidence). The evidence from Xu 2023 B suggests that levonorgestrel IUS may result in little to no difference in pregnancy rate compared to oral progestin (147 more pregnant women per 1000, 95% CI from 127 fewer to 547 more; low‐certainty evidence; Table 4).

The cumulative one‐year pregnancy rates after complete pathological response were 40.7% and 37.5%, in the megestrol acetate and levonorgestrel IUS groups, respectively.

Orbo 2014 did not report any data for this outcome.

2.9. Surgery for persistent/progressive disease (hysterectomy)

Orbo 2014 did not report data on the need for surgical treatment.

Xu 2023 B reported that two women underwent hysterectomy. The information was reported globally, without information on the group.

3. Oral progestin plus levonorgestrel intrauterine system compared to oral progestin

Two RCTs evaluated this comparison (Xu 2023 A; Xu 2023 B). Xu 2023 A included women with endometrial cancer; Xu 2023 B included women with AEH. Given the diversity of the studies, we decided not to pool the data for meta‐analysis. Thus, we describe the results narratively, and we present data in Table 4 and Table 3.

In Xu 2023 B and Xu 2023 A all women underwent co‐intervention (hysteroscopic resection of the lesions). Since co‐interventions are balanced in all study arms, we consider that there is no indirectness issue.

Primary outcomes

3.1. Overall survival

This outcome was not evaluated in any study.

3.2. Live birth rate

Xu 2023 A included 63 women with grade 1 endometrioid endometrial carcinoma without myometrial invasion aged 18–45 years old. The study includes two groups:

• megestrol acetate: 31 women received continuous (long‐term, noncyclical) megestrol acetate (160 mg, orally, daily);

• megestrol acetate plus levonorgestrel IUS: 32 women received megestrol acetate (160 mg, orally, daily) combined with levonorgestrel IUS.

All the women in Xu 2023 A had complete hysteroscopic evaluation and resection of lesions before the initiation of treatment.

Amongst the 56 women who achieved complete pathological response, 33 planned for parenthood. In the megestrol acetate group, 5/21 (23.8%) women gave birth to a live newborn. In the group megestrol acetate plus levonorgestrel IUS, 4/12 (33.3%) had a live birth (RR 1.40, 95% CI 0.46 to 4.24; 1 RCT, 33 women; low‐certainty evidence; Analysis 3.5). The evidence from Xu 2023 A suggests that oral progestin plus levonorgestrel IUS may result in little to no difference in live birth rate compared to oral progestin. The absolute effects are 95 more live births per 1000 (95% CI from 129 fewer to 771 more live births). The overall certainty of the evidence is low, downgraded two levels due to imprecision (Table 4; Table 3).

3.5. Analysis.

Comparison 3: Oral progestin plus levonorgestrel intrauterine system compared to oral progestin, Outcome 5: Oral progestin plus levonorgestrel intrauterine system compared to oral progestin. (EC) Live birth rate

The live birth rate was 27.3% (9/33) in total.

Xu 2023 B is a phase II RCT that included 180 women with AEH; 120 women were treated with megestrol acetate or levonorgestrel IUS plus megestrol acetate (60 in each group). All women underwent hysteroscopic resection before treatment. Forty‐seven women planned for parenthood. After achieving a complete pathological response, in the megestrol acetate group, 5/18 women (27.8%) had a live birth. In the megestrol acetate plus levonorgestrel IUS, 5 /13 (38.5%) had a live newborn, (RR 1.38, 95% CI 0.50 to 3.82; 1 RCT, 31 women; low‐certainty evidence; Analysis 3.1). The evidence from Xu 2023 B shows that oral progestin plus levonorgestrel IUS may result in little to no difference in live births compared to oral progestin. The absolute effects are 106 more live births per 1000 (95% CI from 139 fewer to 783 more live births). The overall certainty of the evidence is low, downgraded two levels due to imprecision (Table 4; Table 3).

3.1. Analysis.

Comparison 3: Oral progestin plus levonorgestrel intrauterine system compared to oral progestin, Outcome 1: Oral progestin plus levonorgestrel intrauterine system compared to oral progestin. Live birth rate.

Secondary outcomes

3.3. Progression‐free survival

This outcome was not evaluated in any study.

3.4. Complete pathological response rate

Xu 2023 A included 63 women with grade 1 endometrioid endometrial carcinoma. All the women underwent complete hysteroscopic evaluation and resection of lesions before the initiation of treatment. In the megestrol acetate group, 5/26 women showed complete pathological response at 16 weeks (19.2%), and 16/28 (61.5%) at 32 weeks. In the megestrol acetate plus levonorgestrel IUS group, 7/28 (25.0 %) showed complete pathological response at 16 weeks and 16/26 (57.1%) at 32 weeks. No differences were found between the treatment groups at 16 weeks (RR 1.30, 95% CI 0.47 to 3.59, P = 0.610; 1 RCT, 54 women; low‐certainty evidence; Analysis 3.6) or 32 weeks (P = 0.743). Evidence from Xu 2023 A shows that oral progestin plus levonorgestrel IUS may result in little to no difference in complete pathological response at 16 weeks compared to oral progestin. The absolute effects were 58 more complete pathological response cases per 1000 (95% CI from 102 fewer to 498 more complete pathological responses). The overall certainty of the evidence is low, downgraded two levels due to imprecision (Table 4; Table 3).

3.6. Analysis.

Comparison 3: Oral progestin plus levonorgestrel intrauterine system compared to oral progestin, Outcome 6: Oral progestin plus levonorgestrel intrauterine system compared to oral progestin.(EC) Complete pathological response rate

Xu 2023 B included 180 women with AEH in three groups of 60 women each. The women underwent hysteroscopic resection of the lesions before starting treatment. In the follow‐up, hysteroscopy was performed every three months, and lesions removed, if present. At 16 weeks, 11/43 women in the megestrol acetate group achieved complete pathological response and at 32 weeks, 36/46 women achieved complete pathological response. In the megestrol acetate plus levonorgestrel IUS group, 16/43 women achieved complete pathological response at 16 weeks (RR 1.45, 95% CI 0.77 to 2.76; 1 RCT, 86 women; low‐certainty evidence; Analysis 3.2), and 38/48 women at 32 weeks. Evidence from Xu 2023 B shows that oral progestin plus levonorgestrel IUS may result in little to no difference in complete pathological response at 16 weeks compared to oral progestin. The absolute effects are 115 more complete pathological response cases per 1000 (95% CI 59 fewer to 450 more complete pathological response cases). The overall certainty of the evidence is low, downgraded two levels due to imprecision (Table 4; Table 3).

3.2. Analysis.

Comparison 3: Oral progestin plus levonorgestrel intrauterine system compared to oral progestin, Outcome 2: Oral progestin plus levonorgestrel intrauterine system compared to oral progestin. Complete pathological response rate

3.5. Severe adverse events

Xu 2023 A reported no treatment‐related deaths or serious adverse events (grade 4) during the study. The only grade 3 adverse event was weight gain. Grade 3 weight gain was 2/31 (6.5%) cases in the megestrol acetate group, and 2/28 (7.1%) cases in the megestrol acetate plus levonorgestrel IUS group. The evidence suggests that oral progestin plus levonorgestrel IUS may result in little to no difference in severe adverse effects grade 3 (weight gain) compared to oral progestin (RR 1.11, 95% CI 0.17 to 7.34; 1 RCT, 59 women; low‐certainty evidence; Analysis 3.7). The absolute effects were seven more weight gain cases per 1000 (95% CI from 54 fewer to 409 more weight gain cases). The overall certainty of the evidence is low, downgraded two levels due to imprecision (Table 4; Table 3).

3.7. Analysis.

Comparison 3: Oral progestin plus levonorgestrel intrauterine system compared to oral progestin, Outcome 7: Oral progestin plus levonorgestrel intrauterine system compared to oral progestin.(EC). Severe adverse event (weight gain)

The most common side effects (grade 1‐2) were increased nocturnal urine, decreased libido, weight gain, fatigue, night sweats and abdominal distension. Compared with the megestrol acetate group, more women in the megestrol acetate plus levonorgestrel IUS group experienced vaginal haemorrhage (16.1% vs 46.4%; P = 0.012). Conversely, no significance was found in the weight change between the megestrol acetate group (median, 2.5 kg; 95% CI −3.8 to 19.0) and the megestrol acetate plus levonorgestrel IUS group (median, 3.0 kg; 95% CI −15 to 18.4; P = 0.495).

Xu 2023 B reported no treatment‐related deaths or serious adverse events (grade 4) during the study. No difference in weight change was found between the megestrol acetate plus levonorgestrel IUS group (median, 5.0 kg; IQR, 0.0 to 10.0, P = 0.798) and the megestrol acetate group. Grade 3 weight gain in the megestrol acetate group was 10/58 women (17.2%) and in the megestrol acetate plus levonorgestrel IUS group, 9/54 (16.7%). The evidence suggests that oral progestin plus levonorgestrel IUS may result in little to no difference in severe adverse events grade 3 (weight gain) compared to oral progestin (RR 0.97, 95% CI 0.43 to 2.20; 1 RCT, 112 women; low‐certainty evidence; Analysis 3.3). The absolute effects are five fewer weight gain cases per 1000 (95% CI from 98 fewer to 207 more weight gain cases). The overall certainty of the evidence is low, downgraded two levels due to imprecision (Table 4; Table 3).

3.3. Analysis.

Comparison 3: Oral progestin plus levonorgestrel intrauterine system compared to oral progestin, Outcome 3: Oral progestin plus levonorgestrel intrauterine system compared to oral progestin. Severe adverse events (weight gain)

Grade 1‐2 weight gain was 25/58 cases (43.1 %) in the megestrol acetate group. In the megestrol acetate plus levonorgestrel IUS group, grade 1‐2 weight gain was 24/54 (44.4%; P = 1.00).

The megestrol acetate group experienced adverse effects similar to those of the megestrol acetate plus levonorgestrel IUS group. Vaginal haemorrhage occurred more often in the megestrol acetate plus levonorgestrel IUS group than in the megestrol acetate group (46.3% vs 19.0%; P = 0.002).

Apart from weight gain, the most frequent adverse events found (all of them grade 1‐2) were:

• increased nocturnal urine: grade 1–2, 29/58 (50.0%) in the megestrol acetate group versus 30/54 (55.6%; P = 0.576) in the megestrol acetate plus levonorgestrel IUS group;

• night sweats: grade 1–2, 23/58 (39.7%) in the megestrol acetate group versus 22/54 (40.7%) in the megestrol acetate plus levonorgestrel IUS group (P= 1.000);

• insomnia grade 1–2, 21/58 (36.2%) in the megestrol acetate group versus 18/54 (33.3%; P = 0.843) in the megestrol acetate group

• libido decreased by grade 1–2, 21/58 (36.2%) in the megestrol acetate group compared to 23/54 (42.6%) in the megestrol acetate plus LNG group (P = 0.563).

3.6. Quality of life

This outcome was not evaluated in any study.

3.7. Psychological symptoms

This outcome was not evaluated in any study.

3.8. Pregnancy rate

Of the 56 women in Xu 2023 A who achieved complete pathological response, 33 women had recent plans for parenthood. The cumulative one‐year pregnancy rate after complete pathological response was 33.3% in the megestrol acetate group and 33.3% in the megestrol acetate plus levonorgestrel IUS group (log‐rank test P = 0.698). In the megestrol acetate group, 13/21 women (61.9%) became pregnant. In the megestrol acetate plus levonorgestrel IUS group, 9/12 (75%) women became pregnant (RR 1.21, 95% CI 0.76 to 1.94; 1 RCT; 33 women; Table 4).

Of the 166 women in Xu 2023 B who achieved complete pathological response, 47 women planned for parenthood. The pregnancy rate was 76.6% (36/47) in total, 66.7% (12/18) in the megestrol acetate group, and 84.6% (11/13) in the megestrol acetate plus levonorgestrel IUS group (RR 1.27, 96% CI 0.85 to 1.89; 1 RCT; 31 women; Table 4).

3.9. Surgery for persistent/progressive disease (hysterectomy)

Xu 2023 A: 2/20 women underwent hysterectomy for persistent disease, and 1/31 underwent surgery for relapse of the lesions. None of the women experienced progression of the disease during treatment. In the megestrol acetate group, 1/30 women underwent surgery. In the megestrol acetate plus levonorgestrel IUS group, 2/29 women underwent surgery. The evidence suggests that oral progestin plus levonorgestrel IUS may result in little to no difference in surgery for persistent/progressive disease compared to oral progestin (RR 2.07, 95% CI 0.20 to 21.60; 1 RCT, 59 women; low‐certainty evidence; Analysis 3.8). The absolute effects were 36 more women who needed surgery per 1000 (95% CI from 27 fewer to 687 more women who needed surgery). The overall certainty of the evidence is low, downgraded two levels due to imprecision (Table 4; Table 3).

3.8. Analysis.

Comparison 3: Oral progestin plus levonorgestrel intrauterine system compared to oral progestin, Outcome 8: Oral progestin plus levonorgestrel intrauterine system compared to oral progestin.(EC) Surgery for persistent/progressive disease (Hysterectomy)

Xu 2023 B: at the time of the last follow‐up, two women underwent hysterectomy: 1/43 in the megestrol acetate group and 1/43 in the megestrol acetate plus levonorgestrel IUS group (RR 1.00, 95% CI 0.06 to 15.48; 1 RCT, 86 women; low‐certainty evidence; Analysis 3.4). The evidence suggests that oral progestin plus levonorgestrel IUS may result in little to no difference in surgery for persistent/progressive disease compared to oral progestin. The absolute effects were 0 fewer women who needed surgery per 1000 (95% CI from 22 fewer to 335 more women who needed surgery). The overall certainty of the evidence is low, downgraded two levels due to imprecision (Table 4; Table 3).

3.4. Analysis.

Comparison 3: Oral progestin plus levonorgestrel intrauterine system compared to oral progestin, Outcome 4: Oral progestin plus levonorgestrel intrauterine system compared to oral progestin. Surgery for persistent/progressive disease (hysterectomy)

None of the women experienced progressive disease during the treatment.

4. Medroxyprogesterone acetate compared to megestrol acetate

We have included two studies for the comparison between medroxyprogesterone acetate and megestrol acetate (Chen 2016; Park 2013). Both studies are retrospective, observational NRSs, and we describe the results narratively. Chen 2016 included women with AEH and endometrial cancer. Park 2013 included only women with endometrial cancer.

Primary outcomes

4.1. Overall survival

This outcome was not evaluated in any study.

4.2. Live birth rate

Chen 2016 reported that, of 33 women with fertility desire, 11 delivered full‐term live babies (6/16 women with AEH, and 5/37 in women with endometrial cancer). These data were reported globally, not by treatment group (Table 4).

Park 2013 did not evaluate this outcome.

Secondary outcomes

4.3. Progression‐free survival

This outcome was not evaluated in any study.

4.4. Complete pathological response rate

Chen 2016 found similar complete pathological response rates with medroxyprogesterone acetate (23/32 women (71.8%)) as with megestrol acetate (17/21 women (80.9%); RR 0.89, 95% CI 0.66 to 1.20). Similarly, Park 2013 did not report differences, with a complete pathological response of 70/91 women (76.9%) in the medroxyprogesterone acetate group and 45/57 (78.9%) in the megestrol acetate group (RR 0.97, 95% CI 0.82 to 1.16; Table 4).

4.5. Severe adverse events

No severe adverse events (grade 3+) were found.

Chen 2016 reported results on adverse events globally but not by treatment group. Park 2013 did not evaluate this outcome.

4.6. Quality of life

This outcome was not evaluated in any study.

4.7. Psychological symptoms

This outcome was not evaluated in any study.

4.8. Pregnancy rate

Both studies reported the data globally, not by treatment group (Chen 2016; Park 2013).

4.9. Surgery for persistent/progressive disease (hysterectomy)

Both studies reported the data globally, not by treatment group (Chen 2016; Park 2013).

5. Progestin high dose compared to progestin low dose

Primary outcomes

Two studies reported results for this comparison (Park 2013; Simpson 2014). The two studies are retrospective, observational NRSs. We describe their results narratively. Park 2013 evaluated endometrial cancer; Simpson 2014 included women with AEH and endometrial cancer.

5.1. Overall survival

Neither of the studies evaluated this outcome.

5.2. Live birth rate

Neither of the studies evaluated this outcome.

Secondary outcomes

5.3. Progression‐free survival

Neither of the studies evaluated this outcome.

5.4. Complete pathological response rate

Park 2013 included women with endometrial cancer treated with medroxyprogesterone acetate, with a mean dose of 500 mg/day (30 mg/day to 1500 mg/day) or megestrol acetate with a mean dose of 160 mg/day (40 mg/day to 240 mg/day). Simpson 2014 included women with AEH and endometrial cancer, treated with high‐dose progestin (medroxyprogesterone acetate > 100 mg/day or megestrol acetate > 80 mg/day) or low‐dose progestin (medroxyprogesterone acetate < 100 mg/day or megestrol acetate < 80 mg/day).

Park 2013 found that progestin dose was not associated with complete pathological response: 59/75 (78.6%) with high‐dose progestin versus 56/73 (76.7%) with low‐dose progestin (RR 1.03, 95% CI 0.86 to 1.22). Simpson 2014 did not find any differences in the time to reach a complete pathological response with high‐dose progestins (HR 0.92, 95% CI 0.36 to 2.33; Table 4).

5.5. Severe adverse events

None of the studies evaluated this outcome.

5.6. Quality of life

None of the studies evaluated this outcome.

5.7. Psychological symptoms

None of the studies evaluated this outcome.

5.8. Pregnancy rate

Park 2013 reported data globally but not by treatment group.

Simpson 2014 did not evaluate this outcome.

5.9. Surgery for persistent/progressive disease (hysterectomy)

Park 2013 reported data globally, not by treatment group.

Simpson 2014 analysed globally, not by treatment group.

6. Gonadotropin‐releasing hormone agonist (GnRHa) combined with levonorgestrel IUS or letrozole

Only the single‐centre retrospective cohort study by Zhou 2017 performed this comparison. See Table 4.

Primary outcomes

6.1. Overall survival

No data on this outcome were reported.

6.2. Live birth rate

No data on this outcome were reported.

Secondary outcomes

6.3. Progression‐free survival

No data on this outcome were reported.

6.4. Complete pathological response rate

This effect was selected in many analyses as indicative of the intervention‐outcome relationship. No differences were observed in complete pathological response (HR 0.957, 95% CI 0.850 to 1.077); no raw data were available (Table 4).

6.5. Severe adverse events

No data on this outcome were reported.

6.6. Quality of life

No data on this outcome were reported.

6.7. Psychological symptoms

No data on this outcome were reported.

6.8. Pregnancy rate

No data on this outcome were reported.

6.9. Surgery for persistent/progressive disease (hysterectomy)

No data on this outcome were reported.

7. Endometrial curettage with or without oral progestin

The study by Lindahl 1991 gave rise to three publications. An RCT first published in 1991, including 292 women with endometrial hyperplasia; of them, 26 women suffered from atypical hyperplasia. The women were randomised into treatment groups either with only endometrial curettage (endometrial abrasio as reported by the study authors) or with curettage plus medroxyprogesterone acetate. In 1994, two further papers were published on the prospective follow‐up of the same and other women, with each group being included in one of these two different publications (Table 4).

Primary outcomes

7.1. Overall survival

No data on this outcome were reported.

7.2. Live birth rate

No data on this outcome was reported.

Secondary outcomes

7.3. Progression‐free survival

This outcome was not reported.

Lindahl 1991 described a similar progression from endometrial curettage plus medroxyprogesterone acetate (11/11) to curettage alone (14/15) (RR 1.07, 95% CI 0.94 to 1.23) in women with atypical hyperplasia after three months.

7.4. Complete pathological response rate

Lindahl 1991 suggests that adding medroxyprogesterone acetate to endometrial curettage may increase the complete pathological response rate (9/10) compared to endometrial curettage alone (6/15) after 9 to 12 months (RR 2.25, 95% CI 1.17 to 4.32; Table 4).

7.5. Severe adverse events

No data on this outcome were reported.

7.6. Quality of life

No data on this outcome were reported.

7.7. Psychological symptoms

No data on this outcome were reported.

7.8. Pregnancy rate

No data on this outcome was reported.

7.9. Surgery for persistent/progressive disease (hysterectomy)

Lindahl 1991 suggests a higher proportion of hysterectomies in women with atypical hyperplasia treated with curettage plus medroxyprogesterone acetate (3/9, 33.3%) than those treated with curettage alone (3/15, 20%) five years after the initial treatment (RR 1.67, 95% CI 0.42 to 6.56; Table 4).

8. Hysteroscopic resection compared to oral progestin for atypical endometrial hyperplasia and endometrial cancer

We did not identify any studies that reported this comparison and no data are available. We did not produce a summary of findings table for this outcome because no studies were included. This fact does not mean that this comparison is not clinically relevant.

9. Oral progestin plus levonorgestrel IUS compared to levonorgestrel IUS

One RCT studied this comparison in 180 women with AEH (Xu 2023 B). The study included three groups, each with 60 women. All the women underwent the co‐intervention, hysteroscopic resection of the lesions. Since the co‐interventions are balanced in all study arms, we consider that there is no indirectness issue.

Primary outcomes

9.1. Overall survival

This outcome was not evaluated.

9.2. Live birth rate

Of the 180 women with AEH included in Xu 2023 B, 47 women planned for parenthood. In two of the study arms, 120 women were treated with megestrol acetate plus levonorgestrel IUS or levonorgestrel IUS (60 in each group). After achieving complete pathological response, in the levonorgestrel IUS group, 8/16 women (50%) had a live birth. In the megestrol acetate plus levonorgestrel IUS, 5/13 (38.5%) women had a live newborn (Table 4).

Secondary outcomes

9.3. Progression‐free survival

This outcome was not evaluated.

9.4. Complete pathological response rate

At 16 weeks, in the levonorgestrel IUS group, Xu 2023 B reported complete pathological response in 21/46 (35.0%) women, and at 32 weeks in 43/53 (72.0%) women. In the megestrol acetate plus levonorgestrel IUS group, complete pathological response was observed in 16/43 women (29.4%) and 38/48 women (70.3%) at 16 and 32 weeks, respectively. All the women underwent hysteroscopic resection of the lesions before starting the treatment: in the follow‐up, hysteroscopy was performed every three months, and lesions removed, if present (Table 4).

9.5. Severe adverse events

Xu 2023 B did not observe any treatment‐related deaths or serious adverse events (grade 4) during the study. The levonorgestrel IUS group achieved less weight gain (median 0.0 kg; IQR 0.0–10.0). Grade 3 weight gain was 2/60 cases (3.3 %) in the levonorgestrel IUS group, and 9/54 cases (16.7%) in the megestrol acetate plus levonorgestrel IUS group (RR 5.00, 95% CI 1.13 to 22.13; Table 4).

Grade 1‐2 weight gain was 5/60 (8.3%) cases in the levonorgestrel IUS group and 24/54 (44.4%) cases in the megestrol acetate plus levonorgestrel IUS group.

Vaginal haemorrhage occurred in the levonorgestrel IUS group in 16/60 cases (26.7%), and in the megestrol acetate plus levonorgestrel IUS group, in 25/60 cases(46.3%).

Fewer women in the levonorgestrel IUS group experienced increased nocturnal urine, night sweats, insomnia, or facial oedema compared with the megestrol acetate group.

Apart from weight gain, the most frequent adverse events found (all of them grade 1‐2) were:

• increased nocturnal urine, 13/60 (21.7%) in the levonorgestrel IUS group compared to 30/54 (55.6%) in the megestrol acetate plus levonorgestrel IUS group;

• night sweats, 9/60 (15.0%) cases in the levonorgestrel IUS group versus 22/54 (40.7 %)in the megestrol acetate plus levonorgestrel IUS group;

• insomnia, 8/60 (13.3%) cases in the levonorgestrel IUS group versus 18/54 (33.3%) in the megestrol acetate plus levonorgestrel IUS group.

• Libido decreased by a grade of 1–2: 15/60 (25.0%) in the levonorgestrel IUS compared to 23/54 (42.6%) in the megestrol acetate plus LNG group.

9.6. Quality of life

This outcome was not evaluated.

9.7. Psychological symptoms

This outcome was not evaluated.

9.8. Pregnancy rate

Amongst the 166 women in Xu 2023 B who achieved complete pathological response, 47 women planned for parenthood. The pregnancy rate was 76.6% (36/47) in total; 16.7% (9/54) in the levonorgestrel IUS group, and 84.6% (11/13) in the megestrol acetate plus levonorgestrel IUS group (RR 5.08, 95% CI 2.68 to 9.63; Table 4).

The cumulative one‐year pregnancy rates after complete pathological response were 37.5% and 38.5% in the levonorgestrel IUS, and megestrol acetate plus levonorgestrel IUS groups, respectively.

9.9. Surgery for persistent/progressive disease (hysterectomy)

No women underwent surgery (0/58) in the levonorgestrel IUS group, and 1/55 women in the megestrol acetate plus levonorgestrel IUS group underwent surgery (Xu 2023 B).

10. Weight loss plus progestin (any route) compared to progestin (any route)

We did not identify any studies for this comparison and no data are available. We did not produce a summary of findings table because there were no included studies. This fact does not mean that this comparison is not clinically relevant.

Sensitivity analysis

We were unable to conduct a sensitivity analysis because only two studies were included in each comparison suitable for meta‐analysis.

Regarding the subgroup analysis, we planned to do a subgroup analysis by grouping the studies according to histology: endometrial cancer or AEH. We could not carry out a subgroup analysis in the comparison metformin plus progestin compared to progestin (complete pathological response rate). In RCTs, we only had information on AEH, and not on endometrial cancer.

In future updates of this review, with the inclusion of more studies, we will perform sensitivity analysis.

Assessment of heterogeneity

To minimise heterogeneity, we grouped the clinically similar studies together for a common measurement of the outcomes. We assessed heterogeneity by visual inspection and superimposition of CI data. We calculated the I² value in the comparisons. We could not perform a subgroup analysis to compare metformin + progestin versus progestin alone.

Assessment of reporting biases

To minimise the risk of reporting biases, we tried to obtain all the unpublished results. We thoroughly searched in the grey literature, and we checked the citation lists of studies included, textbooks and systematic reviews through hand searching. Our searches included the CENTRAL (Cochrane Library), MEDLINE and Embase databases, and other sources of information (e.g. clinicaltrials.gov amongst others), and included languages other than English. We contacted the principal investigators of ongoing trials or studies awaiting classification for relevant data. No author of an ongoing trial could share information with us.

We did not explore the sources of asymmetry in funnel plots, because we included fewer than 10 RCTs. We qualitatively assessed the reporting biases in all outcomes.

In our review, we conclude that reporting bias is unclear.

Discussion

Summary of main results

The objective of this review was to compare the effectiveness and safety of various fertility‐sparing treatments, including oral progestin, levonorgestrel IUS, metformin, and bariatric or hysteroscopic surgery, for AEH and presumed stage IA grade 1 endometrioid endometrial cancer. We identified 12 eligible studies, comprising six RCTs and six NRSs. Five outcomes relevant to this review were evaluated across several studies: complete response rate, live birth rate, pregnancy rate, severe adverse events, and surgery due to persistent or progressive disease (hysterectomy).

Some outcomes we proposed in our protocol (Fernandez‐Montoli 2018), such as overall survival, progression‐free survival, psychological symptoms, and quality of life, were not addressed in the included studies. Additionally, we found insufficient studies on many proposed interventions (e.g. appetite suppressants, fat malabsorption drugs, serotonin receptor antagonists, and surgical interventions including bariatric surgery or operative hysteroscopy) to include them in our review. Evaluating these interventions and outcomes could provide valuable information for clinicians and women when choosing fertility‐sparing treatments.

In the risk of bias assessment, we classified five RCTs as being at high risk of bias and one study as being at unclear risk of bias (Yang 2020). See Figure 2; Figure 3; Risk of bias in included studies.

Our review included five main comparisons. However, we combined for meta‐analysis only two RCTs in the comparison, metformin plus progestin compared to progestin. Additionally, we identified two NRSs related to this comparison, which we summarised without meta‐analysis (see Table 4). In the comparisons, levonorgestrel IUS compared to oral progestin, and oral progestin plus levonorgestrel IUS compared to oral progestin, we found two RCTs for each comparison. We did not combine them for meta‐analysis but summarised them without meta‐analysis (Table 4). We did not find any studies that compared hysteroscopic resection compared to oral progestin or weight loss plus progestin (any route) compared to progestin (any route). Other comparisons yielded limited studies (both RCTs and NRSs), which we summarised without meta‐analysis (Table 4).

Our findings indicate that metformin plus progestin may have little to no effect on the live birth rate compared to progestin. However, the addition of metformin may slightly increase the complete pathological response rate. Furthermore, metformin plus progestin appears to have little to no effect on the need for surgery for persistent/progressive disease compared to progestin. In Yang 2020, weight gain was the most frequently observed adverse event. Severe adverse events (grade 3‐4) were rare, with weight gain remaining the most frequently seen (Table 1).

When comparing metformin plus progestin to progestin, the evidence suggests that adding metformin may slightly enhance the complete pathological response rate. However, the certainty of the evidence is low, and our confidence in the effect estimate is limited, indicating that the true effect may differ significantly from our estimate (Table 1).

We also observed that levonorgestrel IUS may result in little to no difference in the live birth rate compared to oral progestin (1 RCT in women with AEH) and in the complete pathological response rate (2 RCTs in women with AEH that we did not combine for meta‐analysis). Levonorgestrel IUS may slightly reduce severe adverse events (grade 3; weight gain) compared to oral progestin (1 RCT in women with AEH). Our confidence in this effect estimate is limited due to the low certainty of the evidence (Table 2).

Regarding live birth rate, complete pathological response rate (at 16 weeks), severe adverse events (grade 3, weight gain), and surgery for persistent/progressive disease (hysterectomy). Evidence from two RCTs (one involving women with endometrial cancer and the other women with AEH, not combined for meta‐analysis) suggests that oral progestin plus levonorgestrel IUS may result in little to no difference compared to oral progestin (see Table 3).

The combination of two hormonal treatment routes (oral progestin plus levonorgestrel IUS) may not enhance the complete pathological response compared to oral progestin, with similar live birth rates, severe adverse events, and surgery for persistent/progressive disease. However, the certainty of the evidence is low, and our confidence in the effect estimates is limited. At present, there is insufficient high‐quality data to enable clinicians to identify a single optimal drug, route, dose, or surgical treatment.

Overall completeness and applicability of evidence

We conducted a comprehensive search that identified a variety of interventions, populations, and outcomes. We believe that the 12 studies included in our meta‐analysis are relevant to the questions we are addressing. Given the broad scope of our review, our bibliographic search was also quite complex and broad. This complexity may have affected the precision of our search results.

Initially, the search did not include 'hysteroscopy', but after a peer‐reviewer’s suggestion, we added 'hysteroscopic resection' as a key intervention for future searches.

We excluded studies involving women older than 50 years, those with co‐morbidities, very high BMI, or those not eligible for surgical treatment, as these subpopulations could influence the effects of the interventions. In our protocol (Fernandez‐Montoli 2018), we decided to include NRSs because we anticipated that our PICO (Population, Intervention, Comparison, and Outcome) question would be difficult to answer solely through randomised trials. However, after re‐running the search, we ended up including some RCTs. The results from studies awaiting classification may also change our findings. In future updates, we may modify our inclusion criteria to focus only on RCTs.

We analysed data from participants with AEH and stage IA1 endometrial cancer together when both conditions were present. Due to clinical differences among studies—such as median age and cointerventions like hysteroscopic tumour resection—we did not compare some interventions within a meta‐analysis (e.g. levonorgestrel IUS versus oral progestin) when studies had significant heterogeneity. For the complete pathological response rate, we used data at 16 weeks instead of 32 weeks for consistency (Yang 2020).

The diversity of interventions reflects real‐world practices and suggests that our systematic review has potential external validity, but more standardised studies are needed. Some important outcomes, such as overall survival, progression‐free survival, psychological health, and quality of life, or many proposed interventions, were not sufficiently addressed, highlighting areas for future research.

Quality of the evidence

We identified 12 studies with 904 participants. Due to the characteristics of our review, numerous interventions, doses and follow‐up schedules were carried out, and the studies included small numbers of participants.

We rated the overall bias judgment of the included RCTs according to RoB 1, the Cochrane tool for assessing risk of bias (Higgins 2017); we described the NRSs narratively and did not assess their risk of bias. We classified five RCTs as being at high risk of bias, and one as being at unclear risk of bias.

The results of our overall risk of bias judgements (high risk and unclear risk of bias) do not provide enough evidence to support any intervention with certainty.

We used the GRADE system to evaluate the certainty of evidence for each outcome. GRADE rates evidence from very low to high (Ryan 2016). The baseline rating of high is downgraded according to the results of the risk of bias, inconsistency, indirectness, imprecision or publication bias. The evidence identified in our review is of low certainty, which means our certainty in our conclusions is low.

For the live birth rate, the complete pathological response rate and surgery for persistent/progressive disease, in the comparison metformin plus progestin versus progestin (2 RCTs), we downgraded the levels to low, due to imprecision in all outcomes, arising from the small sample size and wide CI including the no‐effect line. In the complete pathological response, the boundaries of the CI (1.07 to 3.19) do not include the possibility of harm or no‐effect line. For severe adverse effects (grade 3 +), we downgraded two levels to low due to imprecision (Table 1).

In the comparison of the levonorgestrel IUS compared to oral progestin (2 RCTs), for the live birth rate, complete pathological response rate, and severe adverse events, we downgraded two levels to low in all outcomes (2 studies were not combined in a meta‐analysis), due to imprecision, arising from the small sample size and wide CI including the no‐effect line (Table 2).

In the comparison, oral progestin plus levonorgestrel IUS compared to oral progestin for the live birth rate, complete response rate, severe adverse events, and surgery for persistent/progressive disease, we downgraded to low in all outcomes (2 studies were not combined in meta‐analysis), due to imprecision, arising from the small sample size and wide CI including the no‐effect line. Table 3

The most frequent reason for downgrading was imprecision. We found very few studies, with few participants. This is probably the reason why our estimates are not precise enough to determine the effect of the interventions. The addition of more high‐quality studies should improve the precision of future estimates and the robustness of the results.

Our findings concerning the primary and secondary outcomes from the studies included in our analysis are shown in the summary of findings for the main comparisons (Table 1; Table 2; Table 3).

Potential biases in the review process

For this review, we conducted an extensive search of published and unpublished studies, to reduce the risk of unpublished studies not being included. Furthermore, we used a wide range of terms to define the interventions considered in our review.

Two review authors (MEF and NC) screened all the database records, and were doubled by other review authors (CJ, JL, JS, PV). The fact that not all the records were screened by the same two authors could be a cause of bias. However, this separation was necessary, given the large number of database records to be screened. To counteract this, all the review authors shared and discussed the data to ensure that their evaluations were as equitable as possible.

During the screening of titles and abstracts, as well as during the data collection phases, we contacted study authors to request extra information or the full texts needed to evaluate some studies. Some study authors sent us the data needed; in other cases, results were still not available (for ongoing studies and studies awaiting classification) or the study authors did not reply to our emails.

In the studies included in our analysis, we found a different number of interventions and types of studies (RCTs, NRSs). To avoid biased results, we grouped the studies into six main comparisons; we narratively described studies whose statistical combination we considered inappropriate. We were only able to perform meta‐analysis in one comparison with two RCTs, because we did not find enough RCTs that were sufficiently homogeneous for the other comparisons. In addition, the risk of bias of these RCTs does not provide enough evidence to support any intervention with certainty.

To deal with heterogeneity, we used a random‐effects model. We could not carry out a sensitivity analysis because we could only combine two RCTs for meta‐analysis. We were also unable to perform a subgroup analysis, because we did not find RCTs with endometrial cancer to compare with endometrial hyperplasia results. The subgroup analysis would have allowed us to explore the impact of the histology (endometrial cancer/AEH) on the complete pathological response rate to metformin plus progestin or progestin alone.

We included RCTs and NRSs. We analysed RCTs in the comparison metformin plus progestin compared to progestin, while other RCTs and NRSs were described narratively. For this reason, we evaluated the risk of bias using the ROB 1 for RCTs only.

Agreements and disagreements with other studies or reviews

Other studies

Several leading institutions have outlined guidelines for hormone interventions, particularly those involving progestins or IUSs that release progestins (Abu‐Rustum 2025; Morrison 2022; RCOG 2016; Rodolakis 2023). Research is also being conducted on drugs that reverse insulin resistance to enhance survival rates in endometrial cancer; metformin was associated with reversion of AEH and higher overall survival in women with endometrial cancer who used metformin compared to non‐metformin users (Meireles 2017). Feng 2018 also reported a higher complete pathological response rate in women treated with a high dose of metformin compared to those on a low dose. However, this RCT had a small sample size and a high risk of bias. Notably, no severe adverse events (grade 3+) were reported, suggesting that this combination of treatments could represent a promising new avenue for clinical practice.

Regarding the use of levonorgestrel IUS compared to oral progestin, one NRS also reported similar pathological response in AEH (Mandelbaum 2020). Other studies also indicate that levonorgestrel IUS may be related to fewer adverse effects than oral progestins (Apgar 2000; Mittermeier 2020; Shoupe 2016).

Hysteroscopic resection of endometrial cancer as a fertility‐sparing approach has been evaluated, but the existing studies are primarily case series without comparison groups. Research involved women treated with hysteroscopic resection alongside oral progestin or levonorgestrel IUS (Falcone 2017; Giampaolino 2019; Laurelli 2011; Marton 2014; Mazzon 2010; Shan 2013). While these studies report favourable complete pathological response rates, they lack comparative data, making it difficult to evaluate the effectiveness of hysteroscopic resection. Live birth rates reported varied significantly, ranging from 33.3% in Laurelli 2011 to 86.6% in Mazzon 2010, with complete pathological response rates between 78.6% and 100% (Falcone 2017; Giampaolino 2019; Laurelli 2011; Marton 2014; Mazzon 2010; Shan 2013).

Other meta‐analyses

Two previous meta‐analyses evaluate live‐birth rate. One by De Rocco 2022 included 29 studies, including case series, evaluating fertility‐sparing treatments for AEH and well to moderately differentiated endometrial cancer (grade I to II). In this study, live birth rate after using levonorgestrel IUS was reported at 80.8% (95% CI 69.5 to 80). In comparison, the rates were 68.8% (95% CI 56 to 80.3) for medroxyprogesterone acetate or megestrol acetate, 25.9% (95% CI 14.9 to 39) for oral progestin combined with levonorgestrel IUS, and 69.9% (95% CI 56.1 to 82) for metformin plus oral progestin.

Additionally, Chae‐Kim 2021's meta‐analysis of six studies found similar live birth rates between women treated with metformin plus progestin and those who received progestin (odds ratio (OR) 0.46, 95% CI 0.21 to 1.03). This meta‐analysis included six RCTs and NRSs and showed no difference adding metformin to oral progestins (pooled OR for remission rate of 1.35, 95% CI 0.91 to 2.00). However, recurrence rates were lower for the combination of metformin and progestin compared to progestin alone (pooled OR 0.46, 95% CI 0.24 to 0.91).

A meta‐analysis of 24 observational studies indicated that oral progestin had a lower regression rate (69% versus 90%, P = 0.03) compared to levonorgestrel IUS in women with AEH (Gallos 2010). However, Guillon 2019 conducted a meta‐analysis involving 65 studies with 1604 women, concluding that the use of medroxyprogesterone acetate was associated with a non‐significant lower remission probability compared with megestrol acetate, levonorgestrel IUS, and GnRH agonists.

Other meta‐analysis characteristics

Some systematic reviews and meta‐analyses on fertility‐sparing treatments for endometrial cancer and AEH were published before this review (Baker 2012; Fan 2018; Gallos 2012; Guillon 2019; Koskas 2014; Luo 2018; Wei 2017). Our search was more comprehensive than most of these previous reviews, as we also consulted additional sources from grey literature, which identified a total of 9526 references.

These earlier reviews included RCTs, if available, case series, and cohort studies with small sample sizes, encompassing both prospective and retrospective studies. Most of the studies included were case series without comparative groups (Baker 2012; De Rocco 2022; Fan 2018; Gallos 2012; Guillon 2019; Koskas 2014; Wei 2017).

These reviews primarily analysed the estimated effects of a single intervention without comparing different interventions. Exceptions to this were Wei 2017 and Luo 2018, which compared oral progestin to levonorgestrel IUS, and Chae‐Kim 2021, which compared metformin plus progestin to progestin. Some studies included in the meta‐analysis by Wei 2017 were excluded from our review. Luo 2018 included only one study that we also incorporated into our review (Orbo 2014).

In terms of the objectives of previous research, Baker 2012 focused on studying conservative treatment options for non‐surgical candidates with comorbidities. Meanwhile, Luo 2018 aimed to investigate the conservative treatment of AEH. Some systematic reviews discussed hormone treatments primarily focusing on oral progestin, but also including levonorgestrel IUS and hysteroscopic resection of endometrial cancer (Baker 2012; Fan 2018; Gallos 2012; Guillon 2019; Koskas 2014; Luo 2018; Wei 2017).

Three meta‐analyses did not assess the quality of the evidence (Baker 2012; Guillon 2019; Koskas 2014). One study used the Methodological Index for Non‐Randomised Studies (MINORS) checklist for evaluation (Gallos 2012), while another employed a modified 18‐item Delphi checklist (Wei 2017). Additionally, Fan 2018 and Chae‐Kim 2021 used the Ottawa‐Newcastle scale, whereas only Luo 2018 assessed the risk of bias using the recommended Cochrane tool.

Authors' conclusions

Implications for practice.

The rising prevalence of obesity and delayed childbearing are contributing to an increasing number of women seeking fertility‐sparing treatment for endometrial cancer and atypical endometrial hyperplasia (AEH). Most clinical guidelines recommend progestin as the first‐line therapy for both conditions. However, the optimal treatment regimen, including drug choice, dosage, and route of administration, remains uncertain due to the lack of randomised controlled trials (RCTs) supporting specific approaches.

In our review, we found that several fertility‐sparing options have been explored for AEH and early‐stage endometrial cancer, including oral progestins, metformin combined with progestins, and the levonorgestrel intrauterine system (IUS), with or without oral progestins. Fertility‐sparing treatment has been shown to be both feasible and effective in achieving complete pathological response. Combinations such as metformin with progestins, levonorgestrel IUS, and progestins alone or in conjunction, have all proven their usefulness in the treatment of AEH and endometrial cancer.

Current evidence is that adding metformin to progestin therapy may slightly increase the rate of complete pathological response compared to progestin alone. Data from two RCTs focused on AEH, although not included in meta‐analyses, indicate that levonorgestrel IUS may result in little to no difference in complete response compared to oral progestins, although the proportion of response is higher with levonorgestrel IUS. Similarly, two RCTs (one with women with AEH and one with women with endometrial cancer) investigating the combination of oral progestins and levonorgestrel IUS, suggest little to no difference in complete response compared to oral progestin alone, though the combination showed a higher proportion of response.

Other promising treatments, such as hysteroscopic resection, are under investigation. However, no such studies met the inclusion criteria for our review due to the absence of RCTs.

Among women who achieved complete response and attempted conception, the use of metformin with progestins, levonorgestrel IUS, or the combination of oral progestins and levonorgestrel IUS showed little to no difference in live birth rates compared to oral progestins alone. However, the absolute number of live births was higher in these alternative treatment groups.

Xu 2023 A, Xu 2023 B and Yang 2020 evaluated safety outcomes. Grade 3–4 weight gain was the most common adverse event reported at this severity level. Weight gain (grades 1 to 2 and 3 to 4) was also the most frequently observed side effect. No deaths were reported in any study. For metformin combined with progestin or levonorgestrel IUS, data suggest less weight gain compared to oral progestin alone. In the combined treatment of oral progestin and levonorgestrel IUS for AEH or endometrial cancer, the evidence indicates little to no difference in grade 3 adverse effects (weight gain) compared to oral progestin alone.

Our review highlights the limited available data. We included only a few RCTs, and those had small sample sizes. The certainty of the evidence was low across all comparisons and outcomes, primarily due to the small number of participants. Additionally, the short duration of follow‐up limited the ability to assess long‐term outcomes. Therefore, the best treatment option remains unknown.

When a young woman with early‐stage endometrial cancer or AEH seeks fertility‐sparing treatment, a thorough discussion of the potential risks and benefits is essential. Women should be informed of the uncertainty surrounding the optimal treatment, including its safety, effectiveness, and risk of disease progression. Decisions should be made on a case‐by‐case basis through shared decision‐making.

Implications for research.

Enroling large numbers of women is challenging because many patients prefer well‐established treatments, and many important outcomes, such as survival, pregnancy success, quality of life, and patient perspectives, are rarely reported. More large, well‐designed randomised trials are needed to determine the best treatment options.

Currently, there is a wide variety of treatments for early‐stage endometrial cancer or AEH, but few studies directly compare these options. This makes it challenging to determine which treatment is the most effective and safest. However, several ongoing studies are expected to provide new insights in the coming years. Some of this researches the effects of combining metformin with levonorgestrel IUS, while others examine levonorgestrel IUS alone, with medroxyprogesterone acetate, with everolimus, or compare these to hormonal therapy (megestrol acetate). There are also studies on megestrol acetate alone, weight loss, hysteroscopic resection, or bariatric surgery. Additionally, exploring the differences between using progestins alone or combined with GnRH‐a is of interest. These future studies will generate new data and help facilitate indirect comparisons through network meta‐analyses, aiming to identify the best treatment options in terms of effectiveness and safety.

Furthermore, the new molecular classification of endometrial cancer, especially the Proactive Molecular Risk Classifier for Endometrial Carcinoma (ProMisE) classification proposed by Britton 2019, offers valuable information for patients and doctors. It helps accurately assess the risk of recurrence and death (Talhouk 2017). The recent consensus guidelines (ESGO/ESHRE/ESGE) recommend performing the ProMisE molecular profile in all young women with early‐stage endometrial cancer (Rodolakis 2023). This approach allows for more precise selection of cases suitable for fertility‐sparing treatments and is particularly important when comparing different therapies and analysing outcomes like live birth rates, complete pathological responses, and the need for additional surgery in cases of persistent or progressive disease.

Finally, it would be desirable to have data on long‐term follow‐up (e.g. up to 10 years) in order to obtain important information on the outcomes that appear over time, such as relapse, mortality, quality of life and pregnancy outcomes.

History

Protocol first published: Issue 8, 2018

Appendices

Appendix 1. Search strategies: MEDLINE, Embase and CENTRAL

MEDLINE search strategy

1. exp Endometrial Neoplasms/
2. Endometrial Hyperplasia/
3. (endometri* adj5 (cancer* or tumor* or tumour* or neoplas* or carcinoma* or malignan* or adenocarcinoma* or hyperplasia*)).ti,ab.
4. 1 or 2 or 3
5. exp Progestins/
6. (progest* or gestagen* or megestrol acetate or megace or medroxyprogesterone acetate or provera).ti,ab.
7. Levonorgestrel/
8. (levonorgestrel or mirena).mp.
9. exp intrauterine devices/
10. exp Antineoplastic Agents, Hormonal/
11. exp Gonadotropin‐Releasing Hormone/
12. (GNRH or GnRH or GnRHa or LHRH or LhRH).ti,ab.
13. ((luteinizing or gonadotropin*) adj5 hormone*).ti,ab.
14. ((luteinizing or gonadotropin*) adj5 (analog* or agonist*)).ti,ab.
15. (decapeptyl or leuprorelin or goserelin or triptorelin or nafarelin or synarel or procrin or zoladex).ti,ab.
16. exp Aromatase Inhibitors/
17. (letrozol anastrazol or exemestane).ti,ab.
18. Metformin/
19. metformin.mp.
20. (hormon* adj5 (therap* or treat*)).ti,ab.
21. exp Appetite Depressants/
22. exp Serotonin Antagonists/
23. ((bariatric* or weight reduc* or gastric*) adj5 (resect* or surg* or band or sleeve or procedure* or bypass*)).ti,ab.
24. (appetite adj3 (suppressant* or depressant*)).ti,ab.
25. (serotonin* adj3 antagonist*).ti,ab.
26. 5 or 6 or 7 or 8 or 9 or 10 or 11 or 12 or 13 or 14 or 15 or 16 or 17 or 18 or 19 or 20 or 21 or 22 or 23 or 24 or 25
27. 4 and 26
28. randomized controlled trial.pt.
29. controlled clinical trial.pt.
30. randomized.ab.
31. placebo.ab.
32. clinical trials as topic.sh.
33. randomly.ab.
34. trial.ti.
35. exp Cohort Studies/
36. (cohort* or prospective* or retrospective*).mp.
37. case series.mp.
38. 28 or 29 or 30 or 31 or 32 or 33 or 34 or 35 or 36 or 37
39. 27 and 38
40. exp animals/ not humans.sh.
41. 39 not 40

Key

mp = title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier
ab = abstract
sh = subject heading
ti = title
pt = publication type

Embase search strategy

1. exp endometrium tumor/ 
2. endometrium hyperplasia/ 
3. (endometri* adj5 (cancer* or tumor* or tumour* or neoplas* or carcinoma* or malignan* or adenocarcinoma* or hyperplasia*)).ti,ab. 
4. 1 or 2 or 3 
5. exp gestagen/ 
6. (progest* or gestagen* or megestrol acetate or megace or medroxyprogesterone acetate or provera).ti,ab. 
7. levonorgestrel/ 
8. (levonorgestrel or mirena).mp. 
9. exp intrauterine contraceptive device/ 
10. exp "antineoplastic hormone agonists and antagonists"/ 
11. exp gonadorelin/ 
12. (GNRH or GnRH or GnRHa or LHRH or LhRH).ti,ab. 
13. ((luteinizing or gonadotropin*) adj5 hormone*).ti,ab. 
14. ((luteinizing or gonadotropin*) adj5 (analog* or agonist*)).ti,ab. 
15. (decapeptyl or leuprorelin or goserelin or triptorelin or nafarelin or synarel or procrin or zoladex).ti,ab. 
16. exp aromatase inhibitor/ 
17. (letrozol anastrazol or exemestane).ti,ab. 
18. metformin/ 
19. metformin.mp. 
20. (hormon* adj5 (therap* or treat*)).ti,ab. 
21. exp anorexigenic agent/ 
22. exp serotonin antagonist/ 
23. ((bariatric* or weight reduc* or gastric*) adj5 (resect* or surg* or band or sleeve or procedure* or bypass*)).ti,ab. 
24. (appetite adj3 (suppressant* or depressant*)).ti,ab. 
25. (serotonin* adj3 antagonist*).ti,ab. 
26. 5 or 6 or 7 or 8 or 9 or 10 or 11 or 12 or 13 or 14 or 15 or 16 or 17 or 18 or 19 or 20 or 21 or 22 or 23 or 24 or 25 
27. 4 and 26 
28. controlled clinical trial/ 
29. crossover procedure/ 
30. double‐blind procedure/ 
31. randomized controlled trial/ 
32. single‐blind procedure/ 
33. random*.mp. 
34. factorial*.mp. 
35. (crossover* or cross over* or cross‐over*).mp. 
36. placebo*.mp. 
37. (double* adj blind*).mp. 
38. (singl* adj blind*).mp. 
39. assign*.mp. 
40. allocat*.mp. 
41. volunteer*.mp. 
42. exp case control study/ 
43. cohort analysis/ 
44. case study/ 
45. prospective study/ 
46. retrospective study/ 
47. (cohort* or prospective* or retrospective*).mp. 
48. ((case adj control*) or (case adj series)).mp. 
49. 28 or 29 or 30 or 31 or 32 or 33 or 34 or 35 or 36 or 37 or 38 or 39 or 40 or 41 or 42 or 43 or 44 or 45 or 46 or 47 or 48 
50. 27 and 49

Key

mp = title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier
ab = abstract
sh = subject headingEnter text here]

CENTRAL search strategy

#1. MeSH descriptor: [Endometrial Neoplasms] explode all trees 
#2. MeSH descriptor: [Endometrial Hyperplasia] this term only 
#3. (endometri*) near5 (cancer* or tumor* or tumour* or neoplas* or carcinoma* or malignan* or adenocarcinoma* or hyperplasia*) 
#4. #1 or #2 or #3 
#5. MeSH descriptor: [Progestins] explode all trees 
#6. (progest* or gestagen* or megestrol acetate or megace or medroxyprogesterone acetate or provera) 
#7. MeSH descriptor: [Levonorgestrel] this term only 
#8. (levonorgestrel or mirena) 
#9. MeSH descriptor: [Intrauterine Devices] explode all trees 
#10. MeSH descriptor: [Antineoplastic Agents, Hormonal] explode all trees 
#11. MeSH descriptor: [Gonadotropin‐Releasing Hormone] explode all trees 
#12. (GNRH or GnRH or GnRHa or LHRH or LhRH)
#13. luteinizing or gonadotropin* near5 hormone*
#14. (luteinizing or gonadotropin*) near5 (analog* or agonist*)
#15. (decapeptyl or leuprorelin or goserelin or triptorelin or nafarelin or synarel or procrin or zoladex) 
#16. MeSH descriptor: [Aromatase Inhibitors] explode all trees 
#17. (letrozol anastrazol or exemestane) 
#18. MeSH descriptor: [Metformin] this term only 
#19. metformin 
#20. (hormon*) near5 (therap* or treat*) 
#21. MeSH descriptor: [Appetite Depressants] explode all trees 
#22. MeSH descriptor: [Serotonin Antagonists] explode all trees 
#23. (bariatric* or weight reduc* or gastric*) near5 (resect* or surg* or band or sleeve or procedure* or bypass*)
#24. (appetite) near3 (suppressant* or depressant*) 
#25. (serotonin*) near3 (antagonist*)
#26. #5 or #6 or #7 or #8 or #9 or #10 or #11 or #12 or #13 or #14 or #15 or #16 or #17 or #18 or #19 or #20 or #21 or #22 or #23 or #24 or #25
#27. #4 and #26

Data and analyses

Comparison 1. Metformin plus progestin compared to progestin.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Metformin plus progestin compared to progestin. Live birth rate. RCTs	2	72	Risk Ratio (IV, Random, 95% CI)	1.80 [0.88, 3.68]
1.2 Metformin plus progestin compared to progestin. Complete pathological response rate (12‐16 weeks). RCTs	2	141	Risk Ratio (IV, Random, 95% CI)	1.85 [1.07, 3.19]
1.3 Metformin plus progestin compared to progestin. Pregnancy rate. RCTs	2	72	Risk Ratio (IV, Random, 95% CI)	1.05 [0.65, 1.68]
1.4 Metformin plus progestin compared to progestin. Surgery for persistent/progressive disease (hysterectomy). RCTs	2	166	Risk Ratio (IV, Random, 95% CI)	0.96 [0.24, 3.78]
1.5 Metformin plus progestin compared to progestin. Severe adverse events grade 3‐4.	1	150	Risk Ratio (IV, Random, 95% CI)	0.39 [0.08, 1.95]

Comparison 2. Levonorgestrel intrauterine system (LNG‐IUS) compared to oral progestin.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 Levonorgestrel intrauterine system compared to oral progestin. Live birth rate	1	34	Risk Ratio (IV, Random, 95% CI)	1.80 [0.74, 4.39]
2.2 Levonorgestrel intrauterine system compared to oral progestin.Complete pathological response rate.	1	89	Risk Ratio (IV, Random, 95% CI)	1.78 [0.98, 3.25]
2.3 Levonorgestrel intrauterine system compared to oral progestin.Severe adverse events (weight gain)	1	118	Risk Ratio (IV, Random, 95% CI)	0.19 [0.04, 0.84]
2.4 Levonorgestrel intrauterine system compared to oral progestin. Complete pathological response rate.	1	19	Risk Ratio (IV, Random, 95% CI)	1.24 [0.86, 1.78]

Comparison 3. Oral progestin plus levonorgestrel intrauterine system compared to oral progestin.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
3.1 Oral progestin plus levonorgestrel intrauterine system compared to oral progestin. Live birth rate.	1	31	Risk Ratio (IV, Random, 95% CI)	1.38 [0.50, 3.82]
3.2 Oral progestin plus levonorgestrel intrauterine system compared to oral progestin. Complete pathological response rate	1	86	Risk Ratio (M‐H, Random, 95% CI)	1.45 [0.77, 2.76]
3.3 Oral progestin plus levonorgestrel intrauterine system compared to oral progestin. Severe adverse events (weight gain)	1	112	Risk Ratio (IV, Random, 95% CI)	0.97 [0.43, 2.20]
3.4 Oral progestin plus levonorgestrel intrauterine system compared to oral progestin. Surgery for persistent/progressive disease (hysterectomy)	1	86	Risk Ratio (IV, Random, 95% CI)	1.00 [0.06, 15.48]
3.5 Oral progestin plus levonorgestrel intrauterine system compared to oral progestin. (EC) Live birth rate	1	33	Risk Ratio (IV, Random, 95% CI)	1.40 [0.46, 4.24]
3.6 Oral progestin plus levonorgestrel intrauterine system compared to oral progestin.(EC) Complete pathological response rate	1	54	Risk Ratio (IV, Random, 95% CI)	1.30 [0.47, 3.59]
3.7 Oral progestin plus levonorgestrel intrauterine system compared to oral progestin.(EC). Severe adverse event (weight gain)	1	59	Risk Ratio (IV, Random, 95% CI)	1.11 [0.17, 7.34]
3.8 Oral progestin plus levonorgestrel intrauterine system compared to oral progestin.(EC) Surgery for persistent/progressive disease (Hysterectomy)	1	59	Risk Ratio (IV, Random, 95% CI)	2.07 [0.20, 21.60]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Acosta‐Torres 2020.

Study characteristics
Methods	Retrospective study
Participants	92 patients were included Women 18‐45 years, with fertility desire, with AEH/EIN or Stage IA grade 1 endometrial cancer Intervention group: 34 patients Control group: 58 patients
Interventions	Intervention group: metformin + progestogen Control group: progestogen alone Progestin therapy included one or more of the following: MA at 80 to 160 mg orally daily, MPA at 10 to 40 mg orally daily, prometrium 400 mg orally daily, or LNG‐IUD at 52 mg+ 500‐1000 mg
Outcomes	CR (time to CR was calculated by Kaplan‐Meier analysis), Live birth rate, Recurrence‐free survival, need for surgery (hysterectomy)
Notes	CR was assessed and Kaplan‐Meier analysis used to calculate time to CR. Comparison of potential response predictors was performed with multivariable Cox regression models

Chen 2016.

Study characteristics
Methods	Retrospective cohort study; single‐centre
Participants	53 patients were included Women 20‐42 years‐old, with fertility desire, with AEH or endometrial cancer stage I grade 1 without evidence of pelvic lymph node metastasis or extrauterine disease on Trans ‐ vaginal ultrasound or MRI. Expression of progesterone receptors. Fertility desire Intervention group: 32 patients Control group: 21 patients
Interventions	Intervention group: MPA 250‐500 mg/day, at least 6 months Control group: MA: 160‐480 mg/day, at least 6 months
Outcomes	CR after a median period of 6 (3–24) months. Recurrence‐free survival and disease recurrence. Live birth rate
Notes	AEH and EC are analysed together. The outcome is not the main objective of the study. LNG‐IUS was placed if not childbearing plan. Depending on response GnRH‐a was added for 3‐6 cycles. Financial support: not reported Conflict of interest: not reported

Lindahl 1991.

Study characteristics
Methods	Prospective randomised trial; single‐centre
Participants	323 patients were included Women with endometrial hyperplasia. The study includes glandular‐cystic, adenomatous, and atypical hyperplasia. Pathological assessment according to a classification from Kurman (1994) Intervention group: 15 patients with atypical hyperplasia Control group: 10 patients with atypical hyperplasia
Interventions	After initial diagnostic abrasio, women were randomised to: Intervention group: 500 mg MPA, twice weekly for 3 months followed by regular abrasio after 3, 6, 12 and 24 h after diagnosis. Control group: abrasio only carried out at 3, 9 and 24 months after diagnosis
Outcomes	CR rate. Progression‐free survival (percentages)
Notes	The study has a follow‐up of 3 months. Lindahl 1994 A and B analysed complete response at 9 ‐ 12 and 24 months Financial support: was supported in part by grants from the John and Augusta Persson Foundation for Medical Scientific Research and the University of Lund. Conflict of interest: No information on conflicts of interest is given.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Patients were initially either followed randomly with no further treatment apart from abrasio at regular intervals (after 3 months, 9‐12 months and 2 years), alternatively they were treated with 500 mg medroxyprogesterone acetate twice weekly intramuscularly for 3 months, followed by regular abrasio.
Allocation concealment (selection bias)	Unclear risk	There is no other information on allocation sequence concealment
Blinding of participants and personnel (performance bias) All outcomes	Low risk	There is no information on blinding of participants or personnel but it is likely that there was no blinding of patients or personnel. However the review authors judge that the outcome is not likely to be influenced by lack of blinding.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Pathological assessment according to a classification from Kurman (1994). No information about awareness of the assessors. It is a pathological assessment.
Incomplete outcome data (attrition bias) All outcomes	Low risk	At 3 months there were no missing cases. At 5 years, control, there were 2 missing cases (2/10 cases). Small number of cases included in the study
Selective reporting (reporting bias)	Low risk	Protocol not available, but the reported outcomes are those expected
Other bias	High risk	No information on baseline characteristics of the patients. The study included all types of endometrial hyperplasia and had few cases of AEH

Mitsuhashi 2019.

Study characteristics
Methods	Retrospective cohort study; single‐centre
Participants	86 patients were included Women with fertility desire, with endometrial carcinoma stage IA grade 1 or AEH with fertility desire Intervention group: 63 women (42 EC, 21 AEH) aged 40 years old or less Control group: 23 historical controls with endometrial cancer treated with MPA
Interventions	Intervention group: daily oral 400 mg MPA until remission (6‐12 months) + metformin (initial dose 750 mg/day, increased weekly by 750 mg up to 2250 mg/d) Control group: MPA
Outcomes	Complete response. Live birth rate. Recurrence‐free survival. Adverse events. Response for metabolic profiles
Notes	Intention‐to‐treat analysis. CR at 18 months, 57 months. Control includes only endometrial cancer. Historic cohort with lower BMI, fewer PCOS patients. Also included patients not candidates to metformin Fertility treatment was initiated after 6 cycles of maintenance treatment. Otherwise, a low‐dose Estrogen/progestin or cyclic lower­‐dose progestin was prescribed until the patient wished to conceive Financial support: not reported Conflict of interest: not reported

Orbo 2014.

Study characteristics
Methods	RCT; multicentre (17 centres)
Participants	170 patients were included Women aged 30–70 years with histologically confirmed endometrial hyperplasia according to WHO 94 classification and D‐score. 19 patients presented AEH. Intervention group: 53 patients Control cyclic MPA group: 52 patients Control continous MPA group: 48 patients
Interventions	Intervention group: LNG‐IUD (20 µg/day) 6 months Control group: oral MPA 10 mg /day /10 days cycle, 6 months Control group: oral MPA 10 mg continuous, 6 months
Outcomes	Regression of hyperplasia( 6 months). Adverse effects
Notes	Only some of the participants have AEH. Includes 6% menopausal patients. Fertility is not an objective. The majority of women with endometrial hyperplasia were between 45 and 51 years of age when included in the study Financial Support and conflict of interest: received funding from the Norwegian Cancer Association, the Regional Research Board of Northern Norway (Helse Nord),the Bank of North Norway, and the University of Tromsø. This study also received fees from Bayer for invited lectures given at scientific seminars for gynaecologists (Orbo). For other contributing authors ABV, MA, IP and BS no disclosure of interest exists.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	A computer‐generated list of random numbers was used. A computer random number generator with two strata and fixed block size.
Allocation concealment (selection bias)	Low risk	Central telephone randomisation was used.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Research team was blinded. No blinding information. Since they are using different dose frequency and different treatment routes in each arm, blinding is unlikely. Gynecologists and participants probably were not blinded. However, the review authors judge that the outcome is not likely to be influenced by lack of blinding.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Pathologists were blinded.
Incomplete outcome data (attrition bias) All outcomes	Low risk	17/170 missing. More missing in the continuous MPA group.
Selective reporting (reporting bias)	Low risk	The study was designed according to the CONSORT statement.
Other bias	High risk	The study included all types of endometrial hyperplasia and had few cases of AEH. Also, includes a 6% of postmenopausal patients, and fertility desire wass not specified.

Park 2013.

Study characteristics
Methods	Retrospective multicentric study; multicentre (8 centres)
Participants	148 patients were included Women up to 40 years old, with fertility desire, with uterine endometrioid adenocarcinoma, stage IA, grade 1. No evidence of myometrial invasion on imaging study Intervention group: 57 patients Control group: 91 patients
Interventions	Intervention group: MA 40‐240 mg/day orally (mean 160 mg/day) Control group: MPA 30‐1500 mg/day orally (mean 500 mg/day)
Outcomes	Complete response. Definitive surgical management (hysterectomy). Recurrence‐free survival.
Notes	The median time on treatment was 9.5 months (range 2–53) After treatment, all patients were followed up every 3–6 months using pelvic examination, tumour marker or imaging studies The response to progestin treatment was assessed histologically Financial support: not reported Conflict of interest: not reported

Shan 2014.

Study characteristics
Methods	RCT; single‐centre
Participants	16 patients were included Women aged ≤ 45 years, with fertility desire, with AEH and desire for fertility conservation and met at least one metabolic syndrome criterion Intervention group: 8 patients Control group: 8 patients
Interventions	Intervention group: 160 mg MA daily + 0.5g metformin 3 x d, 12 weeks Control group: 160 mg MA daily, 12 weeks
Outcomes	Complete pathologic response at 12 weeks. Pregnancy rate. Need for surgery (hysterectomy). Severe adverse events
Notes	Pilot study with few cases Financial support: not reported Conflict of interest: not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomised clinical trial. Not reported type of randomisation. There is unclear information on the type of study. in the article: "This was a controlled single‐blinded prospective cohort study" "Patients were randomised into two groups". We have contacted the authors to clarify the information.
Allocation concealment (selection bias)	Unclear risk	No information if participants were aware of their assignment.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	No information on carers' and people delivering the interventions' awareness on assignments of participants intervention. No blinding information, since they are using different dose frequency in each arm, blinding is unlikely. However, the review authors judge that the outcome is not likely to be influenced by lack of blinding.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Not reported if there is blinding of outcome assessment. Pathological diagnosis was confirmed by two experienced gynaecologic pathologists If their opinion differed, a seminar was held in the Department for the final diagnosis.
Incomplete outcome data (attrition bias) All outcomes	High risk	6/22 exclusions before or during treatment (lost or incomplete data) 3/16 lost during follow up (all cases not CR at first). Not specified to which group missing cases belong. 27% missing cases
Selective reporting (reporting bias)	Low risk	The data were analysed according to a prespecified analysis plan.
Other bias	High risk	Small sample size. It was a pilot study The study was performed in a selected subpopulation (patients with AEH and at least one metabolic syndrome criterion).

Simpson 2014.

Study characteristics
Methods	Retrospective study; multicentere (2 centres)
Participants	44 patients were included Women, with fertility desire, with complex atypical hyperplasia or IA G1EC under the age of 45 years treated with oral progestin for the purpose of fertility‐sparing treatment Clinical databases from the Princess Margaret Hospital and Odette Cancer Center in Toronto, Canada, from 2000 to 2011 Intervention group: 11 patients Control group: 33 patients
Interventions	Intervention group: high‐dose (26%) ≥ 100 mg MPA or ≥ 80 mg MA daily Control group: low‐dose (74%) During a median time of 9.5 months (range 2‐53)
Outcomes	CR. Live birth rate. Recurrence‐free survival. Need for surgery (hysterectomy)
Notes	Median follow‐up 39 months (5‐128) Hysterectomy was carried out in 55.8% of cases because of persistent disease 95% had discontinued progestin treatment by their last follow‐up date Financial support: received financial support from The Mount Sinai Hospital/University Health Network Research Grant. Conflict of interest: No conflicts of interest are reported.

Xu 2023 A.

Study characteristics
Methods	Randomised (1:1) clinical trial, unicentre
Participants	56 patients were included Women 18–45 years, with fertility desire, diagnosed with primary EEC were screened, who strongly required fertility‐preserving treatment Intervention group (IUS): 28 patients Control group: 26 patients
Interventions	Intervention group: MA (160 mg oral daily) plus LNG‐IUS Control group: MA group (160 mg oral daily)
Outcomes	Complete response rate within 16 weeks of treatment Secondary endpoints: CR rate within 32 weeks of treatment, severe adverse events, recurrent and pregnancy rate. Live birth rate
Notes	NCT03241914 Financial support: was supported by the National Key Research and Development Program of China (Grant No. 2019YFC1005200 and 2019YFC1005204), Shanghai Medical Centre of Key Programs for Female Reproductive Diseases (Grant No. 2017ZZ010616) and Shen Kang clinical project (SHDC22021219). Conflict of interest: no potential conflict of interest relevant to this article was reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Patients were randomly assigned (1:1) to receive MA or MA + LNG‐IUS by the simple randomisation. Randomisation sequences were prepared according to random‐number tables. Investigators enrolling participants could possibly foresee assignments and thus introduce selection bias, such as allocation based on using an open random allocation schedule (e.g. a list of random numbers).
Allocation concealment (selection bias)	Unclear risk	Investigators enrolling participants could possibly foresee assignments and thus introduce selection bias, such as allocation based on using an open random allocation schedule (e.g. a list of random numbers).
Blinding of participants and personnel (performance bias) All outcomes	Low risk	This study was open label, so that all study physicians and patients were aware of the treatment assignment. However, the pathologists were not aware of the treatment allocations. There was an incomplete blinding, but the review authors judge that the outcome is not likely to be influenced by lack of blinding.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Pathologic diagnosis was confirmed by 2 experienced gynecological pathologists (Dr. Zhu Q and Dr. Zhou XR) according to the World Health Organization pathological classification (2014) [11]. A seminar was held in the pathological department for the final diagnosis if their opinions differed. Pathologic diagnosis was confirmed by 2 masked experienced gynecological pathologists. If there was a disagreement between the 2 pathologists, a seminar was held in the pathological department for the final diagnosis. There is no information on blinding of hysteroscopists.
Incomplete outcome data (attrition bias) All outcomes	Low risk	At 16 weeks( primary outcome CR rate) there were 9 /63 (14.2%) missing cases. Missing cases were relatively high, however, missing outcome data balanced in numbers across intervention groups, with similar reasons for missing data across groups.
Selective reporting (reporting bias)	Low risk	A protocol of the trial was published previously: NCT03241914 Changes in the protocol were specified, and probably did not impact in the results (16 weeks‐32 weeks CR in spite to the originally designed primary and secondary endpoints (3‐ and 6‐month CR rates)
Other bias	High risk	Small sample size (53 patients) (Authors assumed the 16‐week CR rate was 20% in MA group and 30% in MA + LNG‐IUS group; with a power of 0.8 at a 2‐sided significance level of 0.05; requiring an accrual of 458 eligible patients (lost to follow‐up rate < 10%). The sample size did not reach the calculated number; therefore, the power of the test was insufficient. The follow up time was short. The treatment groups were combined with hysteroscopic evaluation and resection of endometrial lesions, which might have concealed the difference in efficacy of the regimens.

Xu 2023 B.

Study characteristics
Methods	Randomized (1:1:1) clinical trial, unicentre
Participants	180 patients were included Women, 18‐45 years, with AEH 60 patients in each group (3 groups).
Interventions	Intervention group: MA (160 mg orally daily) + LNG‐IUS Control group: MA (160 mg orally daily) Control group: LNG‐IUS
Outcomes	Complete response, severe adverse events. Live birth rate. Pregnancy rate
Notes	NCT03241888. It is a communication to congress Financial support: was supported by the National Key Research and Development Program of China (Grant No 2019YFC1005200 and 2019YFC1005204), Shanghai medical Centre of Key Programs for Female Reproductive Diseases (Grant No. 2017ZZ010616), Shanghai sailing program (Grant No. 19YF1404200), and Shen Kang clinical project (SHDC22021219). Conflict of interest:
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomised clinical trial. Simple randomisation. The randomisation sequences were prepared using random number tables. Assignment 1:1:1
Allocation concealment (selection bias)	Unclear risk	The treatment allocation was concealed before the participants were successfully enrolled. The method used to conceal allocation was not specified. None of the clinicians who performed the hysteroscopic evaluation of the patients in this trial or the pathologists who assessed the specimens from this trial were aware of the treatment allocations. The randomisation sequences were prepared using random number tables. Investigators enrolling participants could possibly foresee assignments and thus introduce selection bias, such as allocation based on using an open random allocation schedule (e.g. a list of random numbers).
Blinding of participants and personnel (performance bias) All outcomes	Low risk	This study was open‐labeled, and all patients and study physicians were aware of the treatment assignment. There was no blinding or incomplete blinding, but the review authors judge that the outcome is not likely to be influenced by lack of blinding;
Blinding of outcome assessment (detection bias) All outcomes	Low risk	None of the clinicians who performed the hysteroscopic evaluation of the patients in this trial or the pathologists who assessed the specimens from this trial were aware of the treatment allocations Pathologic diagnosis was confirmed by two experienced gynecological pathologists. If their opinions differed, a seminar was held in the pathology department for the final diagnosis.
Incomplete outcome data (attrition bias) All outcomes	Low risk	For complete response rate, the missing rate was relatively high 26.7% and 15.6% at 16 and 32 weeks of treatment, respectively. However, missing outcome data are balanced in numbers across intervention groups, with similar reasons for missing data across groups.
Selective reporting (reporting bias)	Low risk	A protocol of the trial was published previously: NCT03241888. Changes in the protocol were specified, and probably did not impact in the results (16 weeks‐32 weeks complete response in spite to the originally designed primary and secondary endpoints ( 3‐ and 6‐month CR rates)
Other bias	High risk	All three treatment groups were combined with hysteroscopic evaluation and resection of endometrial lesions, which might have concealed the difference in efficacy of the regimens. The study had a potential source of bias related to the specific study design used. The sample size did not reach the calculated number; therefore, the power of the test was insufficient.

Yang 2020.

Study characteristics
Methods	Randomised clinical trial, unicentre
Participants	150 patients were included Women, 18‐45 years old, with fertility desire, with EEC stage 1A grade 1 or AEH with fertility desire Intervention group: 76 patients Control group: 74 patients
Interventions	Intervention group: metformin (500 mg orally 3 times/day) + MA (160 mg orally, daily) Control group: MA alone 160 mg orally, daily
Outcomes	Complete pathological response rate at 3 months and cumulate CR at 6 months (If endometrium evaluation by hysteroscopy within 16 and 32 weeks of treatment, the evaluation of cumulate CR was at 16 and 32 weeks). Live birth rate. Pregnany rate. Progression‐free survival. Need for surgery (hysterectomy). Severe adverse events
Notes	Clinical trials: NCT01968317 Both groups include hysteroscopic resection 20% missing cases or exclusions in experimental group. In control group 11.9% missing and exclusions Financial support: was funded by the National Key Technology R&D Programme of China (Grant Nos 2019YFC1005200and 2019YFC1005204), National Natural Science Foundation of China (grant nos 81671417 and 81370688), Shanghai Medical Centre of Key Programmes for Female Reproductive Diseases (grant no. 2017ZZ010616), Shanghai Science and Technology Development medical guide project (grant nos 17411961000, 134119a4500, 19411960400) and Municipal Human Resources Development Programme for Outstanding Leaders in Medical Disciplines in Shanghai (grant no. 2017BR035). Conflict of interest: authors declared no conflicts of interest.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomised clinical trial. Computer‐based procedure of simple randomisation (SPSS v 20). 1:1 Before enrolment, treatment assignment of patients was concealed
Allocation concealment (selection bias)	Low risk	Before enrolment, treatment assignment of patients was concealed.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	It was an Open label trial: patients and study physicians were aware of the treatment assignment. Hysterroscopists were not aware of assignment. The review authors judge that the outcome is not likely to be influenced by lack of blinding.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Hysteroscopist was unaware of this trial. There is not information on blinding of radiologists. Pathologists were unaware of the treatment. Pathological diagnosis was Pathological diagnosis was confirmed by two experienced gynaecologic pathologists according to WHO classification (2014).
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	In 16 weeks complete response there were 20% missing cases or exclusions in the control group (progestin group). In the experimental group (metformin plus progestin 11.9%). The reasons for the missing cases were: missing hysteroscopy in both groups. At 32 weeks, missing cases were 7/73 (9.5%) in the progestin group and 5/75 (6.6%) in the metformin plus progestin group. At 12 months, missing cases were: 6/74 and 3/72 ( 8.1 % and 4,1% respectively). The "high risk" judgement in the domain incomplete outcome data is applicable for complete response at 16 weeks. Not applicable to the other outcomes. For this reason we rated this outcome as unclear.
Selective reporting (reporting bias)	Low risk	a protocol of the trial was published previously. NCT01968317 Changes in the protocol were specified, and probably did no impact in the results ( 16 weeks complete response in spite of 12 weeks )
Other bias	Low risk	No other sources of bias detected.

Zhou 2017.

Study characteristics
Methods	Retrospective study; unicentre
Participants	29 patients were included Women younger than 45 years with CAH or early well‐differentiated endometrioid adenocarcinoma of the uterus (EC) stage IA Intervention group: 18 patients Control group: 11 patients
Interventions	Intervention: intramuscular injections of GnRHa every 4 weeks + the LNG‐IUS Control: intramuscular injections of GnRHa every 4 weeks + oral letrozole 2.5 mg daily
Outcomes	Complete response. Length of follow‐up. Pregnancy outcomes
Notes	Median follow‐up 18.7 months (range, 5.6‐54.9 months). Mean age 30.6 (21‐42) Financial support: not reported Conflict of interest: not reported

AEH: atypical endometrial hyperplasia; BMI: body mass index; CAH: complex atypical hyperplasia CR: complete (pathological) response; (E)EC: (endometroid) endometrial cancer; EIN: endometrial intraepithelial neoplasia; EP: estrogen/progestin GnRH‐a: gonadotropin‐releasing hormone agonist; LNG‐IUD/S: levonorgestrel intrauterine device/system; MA: megestrol acetate; MRI: magnetic resonance imaging; MPA: medroxyprogesterone acetate; PCOS: polycystic ovarian syndrome; RCT: randomised controlled trial

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Ayhan 2020	Comparison does not meet protocol. Ineligible study design
Barr 2021	Ineligible patient population
Chung 2019	Comparison does not meet protocol. Ineligible study design
Ciccone 2019	Ineligible study population. Makes reference to a subpopulation with comorbidity that is an exclusion criterion
Dursun 2012	Comparison does not meet protocol. Ineligible study design
Falcone 2017	Comparison does not meet protocol. Ineligible study design
Falcone 2020	Population studied does not meet inclusion criteria. Ineligible patient population
Feng 2018	The main outcome is the study of molecular changes after treatment with metformin. Incomplete information for study design and on methods and patients' characteristics
Gallos 2013a	Population studied does not meet inclusion criteria. Ineligible patient population
Gallos 2013b	Population studied does not meet inclusion criteria. Ineligible patient population
Janda 2021	Population studied does not meet inclusion criteria. Does not evaluate the pregnancy rate
Jobo 2001	Comparison does not meet protocol. Ineligible study design
Kuang 2021	Ineligible study design
Kudesia 2014	Outcomes are not included in our protocol. Ineligible outcomes
Laurelli 2011	Comparison does not meet protocol. Ineligible study design Case‐series study with few patients. The outcome is complete response and the intervention is hysteroscopic resection of endometrial cancer stage IA. Treatment is completed with oral progestogen or levonorgestrel IUS. There is no comparison group, and selection of the treatment depends on the moment of inclusion.
Le Digabel 2006	Comparison does not meet protocol. Ineligible study design
Mandelbaum 2020	Population studied does not meet inclusion criteria. Does not evaluate the pregnancy rate
Marnach 2017	Ineligible patient population. Includes benign hyperplasia
Matsuo 2020	Ineligible patient population
NCT00788671	Ineligible study design
NCT02335203	Ineligible objectives and study design
Novikova 2019	Comparison does not meet protocol. Does not evaluate the pregnancy rate
Raffone 2021	Comparison does not meet protocol. Ineligible study design
Shan 2013	Comparison does not meet protocol. Ineligible study design
Shikeli 2019	Does not compare treatments. Only indicates general relapse rate
Shim 2021	Retrospectively analysed. Does not meet inclusion criteria. It is not a comparative study.
Wang 2014	Comparison does not meet protocol. Ineligible study design
Wang 2019	Comparison does not meet protocol. Ineligible study design
Yamaguchi 2018	Comparison does not meet protocol. Ineligible study design
Yang 2019	It is a communication to conference. Duplicate information
Yuan 2022	Ineligible study design
Zhang 2021	Ineligible patient population. Analyses impact of weight loss on pregnancy only in those women who achieved complete response from a previous fertility‐sparing therapy
Zonglan Xu 2020	Comparison does not meet protocol. Does not evaluate the pregnancy rate

IUS: intrauterine system

Characteristics of studies awaiting classification [ordered by study ID]

Alnemr 2024.

Methods	RCT (open‐label), single‐centre
Participants	148 patients were included Women with AEH who declined hysterectomy Intervention group: 74 patients Control group: 74 patients
Interventions	Intervention group: apply LNG‐IUS Control group: daily oral MA 160 mg Follow‐up by endometrial sampling at 3, 6, 9, 12, 18, and 24 months
Outcomes	Primary outcomes: The success rate and duration until complete regression
Notes

Dong 2023.

Methods	Prospective randomised trial
Participants	Women with EAH or EEC
Interventions	Intervention group: MA (160 mg orally, daily) plus metformin (500 mg orally, three times a day) Control group: MA (160 mg orally, daily)
Outcomes	CR rate, DFS rate and pregnancy rate
Notes

Dong 2024.

Methods	Prospective RCT, single‐centre
Participants	150 patients were included Women with AEH and EEC. From October 2013 to October 2017 Intervention group: 76 patients Control group: 74 patients
Interventions	Intervention group: MA (160 mg orally, daily) plus metformin (500 mg orally, three times a day). Control group: MA (160 mg orally, daily) With a median follow‐up time of 57.7 (26.7, 70.5) months
Outcomes	Time to CR and RFS
Notes

Goh 2023.

Methods	RCT, multicentre
Participants	34 patients were included Women aged 40 years old and below
Interventions	Mirena (LNG‐IUS) compared to megace (MA) in the treatment of AH From January 2020 to January 2023.
Outcomes	Primary outcome: the regression rate Secondary outcomes: side effects, patient acceptability and fertility outcomes
Notes

Goh 2024.

Methods	Phase II RCT, multi‐centre
Participants	36 patients completed the trial. Women who were diagnosed with AEH and between 21 and 40 years old. From January 2020 to January 2024
Interventions	Intervention group: LNG‐IUS Control group: MA (160 mg orally, daily).
Outcomes	Primary outcomes: the regression rates at 3 months, 6 months and 9 months of treatment Secondary outcomes: the side effects, patient acceptability and fertility outcomes
Notes

Liu 2024.

Methods	Prospective randomised trial (open‐label, two‐armed). Multicentre
Participants	226 patients were included
Interventions	Group 1: GnRH‐a with the LNG‐IUS Group 2: GnRH‐a with an aromatase inhibitor (comprising a subcutaneous GnRH‐a injection every 4 weeks and daily oral letrozole 2.5 mg)
Outcomes	Primary outcome: to determine the effectiveness of GnRH‐a‐based re‐treatment in achieving CR at 24 weeks for patients with AEH or EC. Secondary outcomes: assessing the pregnancy rate 12 weeks after treatment, post‐treatment pregnancy outcomes and the rate of recurrence
Notes

NCT05829460.

Methods	Prospective randomised trial
Participants	Women resident in Australia > 18 years diagnosed with cancer of the ovary, cervix, fallopian tubes, placenta, endometrium, vagina or vulva in the previous 60 months, including early, recurrent, advanced or metastatic cancer. (Cancer ‐ Cervical (cervix) Cancer ‐ Other cancer types Cancer ‐ Ovarian and primary peritoneal Cancer ‐ Womb (Uterine or endometrial cancer) Gynaecological cancer; ; Gynaecological cancer )
Interventions	Intervention group: will receive one‐on‐one supervision by an Accredited Exercise Physiologists or Physiotherapists (AEPP). Control group: will receive standard care and will receive general advice about self‐managed exercise, a Fitbit, and the Exercise is Medicine guidelines for gynaecological cancer. The total study period is 24 weeks. It comprises of a 12‐week exercise training intervention and a 12‐week maintenance period.
Outcomes	Primary outcome: the mean difference on the mental component summary (MCS) scores of health‐related quality of life between intervention and control in this trial, as measured by the SF36; the mean difference on the physical component summary (PCS) scores of health‐related quality of life between intervention and control in this trial, as measured by the Short Form‐36 (SF36). Secondary outcomes: balance, blood markers for follicle‐ stimulating hormone, luteinizing hormone, estradiol, progesterone. Blood markers of glycaemic modulation, inflammatory modulation. Body composition, cardiorespiratory fitness, duration of physical activity levels, dynamic upper and lower body muscle strength, exercise self‐efficacy, frequency of physical activity levels, intensity of physical activity levels
Notes

NCT05903131.

Methods
Participants
Interventions	It is the hypothesis that premenopausal women with AEH desire uterine preservation will be more likely to have atypia‐free uterine preservation at one year if they receive progestin in combination with a behavioral weight loss intervention versus progestin plus enhanced usual care.
Outcomes
Notes

NCT06102863.

Methods	Prospective randomised trial
Participants
Interventions	Intervention group: progesterone regimen (oral MPA tablet 250 mg‐500 mg/day or Mirena + GnRHa 3.75 mg subcutaneous injection monthly) combined with statins (oral atorvastatin calcium tablet 20 mg/day; or rosuvastatin 5 mg/day;or pivastatin 2 mg/ day) Control group: progesterone regimen (oral MPA tablet 250 mg‐500 mg/day or Mirena + GnRHa 3.75 mg subcutaneous injection monthly)
Outcomes
Notes

Papakonstantinou 2023.

Methods	Prospective randomised trial, single‐centre
Participants	12 patients were included Women, 32‐38 years old, wishing to preserve fertility, with complex AEH Intervention group 1 (LNG‐IUD alone): 5 women Intervention group 2 (LNG‐IUD + oral progestins): 2 women Control group (oral progestins alone): 5 women
Interventions	Intervention group 1: LNG‐IUD alone Intervention group 2: LNG‐IUD with oral progestins Control group: oral progestins alone
Outcomes	Disease regression, free of disease
Notes

Wang 2024.

Methods	Prospective randomised trial
Participants	60 patients were included Women aged 20‐42 years with AEH and/or grade 1 EAC limited to the endometrium Intervention group: 30 patients Control group: 30 patients
Interventions	Intervention group: MA plus oral metformin (850 mg, twice a day) Control group: MA treatment at the daily dose of 160 mg For at least 6 months. The treatment could extend to 12 months until CR was achieved.
Outcomes	CR, recurrence rate, weight gain
Notes

A(E)H: atypical (endometrial) hyperplasia; CR: complete (pathological) response/remission; DFS: disease‐free survival; EAC: endometriod adenocarcinoma; (E)EC: (endometroid) endometrial cancer; GnRH‐a: gonadotropin‐releasing hormone agonist; LNG‐IUD/S: levonorgestrel intrauterine device/system; MA: megestrol acetate; MPA: medroxyprogesterone acetate; RCT: randomised controlled trial; RFS: recurrence‐free survival

Characteristics of ongoing studies [ordered by study ID]

NCT01943058.

Study name	Megestrol acetate or levonorgestrel‐releasing intrauterine system in treating patients with atypical endometrial hyperplasia or endometrial cancer
Methods	RCT; parallel assignment; open‐label
Participants	Age: 18‐44 years Diagnosis of complex AH or grade 1 EAC of the endometrium diagnosed within 3 months of study enrolment who strongly desire to maintain fertility MRI with myometrial invasion < 50% and not extrauterine metastases
Interventions	Arm A: MA Arm B: LNG‐IUS
Outcomes	Primary outcome: response rate at 6 months Secondary outcome: change in weight. Compliance
Starting date	March 2014
Contact information	Yvonne Lin‐Liu University of Southern California
Notes	Recruitment failure. No results

NCT02342730.

Study name	Weight loss referral for healthier survivorship in obese stage I‐II endometrial cancer survivors or atypical hyperplasia
Methods	Open‐label; single‐arm
Participants	127 women Age: 18‐65 years Inclusion criteria Women with a history of stage I or II EC or a diagnosis of complex AEH BMI ≥ 30 kg/msq
Interventions	Weight loss referral Women are referred to a weight‐loss specialist for assistance with weight loss and medical chart reviews are performed at baseline and every 3 months for 24 months. Women complete Quality‐of‐Life Assessment (EORTC‐ QLQ‐C30 and EORTC‐QLQ‐EN24) at baseline, 12, and 24 months Women are also contacted at 90 days to determine whether they have initiated any weight‐loss interventions.
Outcomes	Primary outcome: accrual with intervention. Compliance with intervention Secondary outcome: ∘weight loss (in kg) at 12 months and 24 months. Compliance with lifestyle changes at 3 months. Incidence of obesity‐related comorbidities. Adverse events. Progression‐free survival. Overal survival. Recurrence rate. Level of functioning, quality of life, and symptomatology
Starting date	17 December 2014
Contact information	Stephanie Blank stephanie.blank@nyumc.org
Notes	Not an RCT. Fertility‐sparing treatment is not a outcome

NCT02397083.

Study name	Levonorgestrel‐releasing intrauterine system with or without everolimus in treatment patients with atypical hyperplasia or stage IA grade 1 endometrial cancer
Methods	RCT; parallel assignment; open‐label
Participants	270 women Age: > 18 years Diagnosis of complex AH or grade 1 EC or focal grade 2 AEC in predominately grade 1 disease EC on endometrial biopsy or dilation and curettage within three months of study enrolment
Interventions	Arm A: LNG‐IUS alone) for 9 months Arm B: LNG‐IUS plus everolimus for 9 months
Outcomes	Primary outcome: response rate at 3 and 6 months of LNG‐IUD alone; response rate at 6 months of LNG‐IUD plus everolimus Secondary outcome: adverse events. Progression‐free survival. Overall survival. Response duration
Starting date	23 September 2015
Contact information	Shannon Westin swesting@mdanderson.org
Notes	The study does not specify if one of the objectives is fertility‐sparing treatment

NCT02990728.

Study name	Mirena® ± metformin as fertility‐preserving treatment for young Asian women with early endometrial cancer
Methods	RCT; parallel assignment; open‐label
Participants	120 participants 20‐40 years old Histologic confirmed grade 1 EAC of the endometrium
Interventions	LNG‐IUS only or LNG‐IUS + metformin
Outcomes	Primary outcome: efficacy of Mirena, with or without metformin Secondary outcome: 1. To discover the morphological and molecular change in the endometrium tumour before and after treatment 2. To discover the effectiveness of adding oral progestin to women who show no good response to assigned 3. To compare the systemic effects 4. Molecular markers and their expression before, during and after treatment
Starting date	March 2016
Contact information	Chang Ting‐Chang; Chang Gung Memorial Hospital
Notes	Estimated primary completion: March 2018

NCT03042897.

Study name	Exercise and diet intervention in promoting weight loss in obese patients with stage I endometrial cancer
Methods	Interventional; single‐group assignment; open‐label
Participants	25 participants Inclusion criteria Clinical stage I EC Low‐grade disease‐positive for estrogens and progesterone receptors BMI ≥ 30 kg/m2
Interventions	Supportive care (exercise and diet) Women undergo aerobic exercise thrice weekly over 95 minutes for up to 16 weeks. Women also undergo multi‐lifestyle interventions based on the DASH diet once weekly over 1 hour for up to 16 weeks
Outcomes	Primary outcome: to determine if participation in a 16‐week exercise and diet intervention would elicit weight loss, assessed by a 10% decrease in fat mass, in EC patients Secondary outcome: to determine if participation in a 16‐week exercise and diet intervention would improve cardiorespiratory fitness (CRF) and muscular strength. Terciary outcome: to determine if participation in a 16‐week exercise and diet intervention could improve quality of life (QOL), fatigue, and depression. To determine the effects of a 16‐week exercise and diet intervention on cardiovascular (CV) health
Starting date	3 February 2017
Contact information	Dieli‐Conwright Christina cdieli@usc.edu
Notes	Estimated primary completion date: 8 September 2020

NCT03463252.

Study name	Value of LNG‐IUS as fertility‐preserving treatment of EAH and EC
Methods	RCT; parallel assignment; open‐label
Participants	224 women For women with EC Inclusion criteria: 1. ≤ 40 years of age 2. Having a strong desire for fertility‐sparing treatment. 3. Histological diagnosis of grade 1 EAC, and the progesterone receptors (PgRs) are positive in immunohistochemistry 4. Disease limited to the endometrium (stage 1A) on MRI For women with AEH Inclusion criteria: 1. ≤ 40 years of age 2. Having a strong desire for fertility‐sparing treatment. 3. Histological diagnosis of AEH
Interventions	For EC Active comparator: MPA 250 mg‐500 mg oral for 3 months per cycle Experimental: MPA + Mirena® for 3 months per cycle Experimental: Mirena® for 3 months per cycle Active comparator: GnRHa + Mirena®. Intramuscular injection 3.75 mg once a month for 3 months per cycle Experimental: Mirena® for 3 months per cycle For AEH Active comparator: Mirena® for 3 months per cycle Experimental: MPA 250 mg‐500 mg oral for 3 months per cycle Active comparator: Mirena® for 3 months per cycle Experimental: GnRHa + Mirena®. Intramuscular injection 3.75 mg once a month for 3 months per cycle
Outcomes	Primary outcome: pathologic response (time frame: 6‐9 months); pregnancy rate (time Fframe: 7‐15 months); live birth rate (time frame: 16‐24 months) Secondary outcome: adverse events (time frame: 1‐9 months)
Starting date	1 April 2018
Contact information	Zheng Ying 93539863@qq.com
Notes	Estimated study completion date December 2020

NCT04008563.

Study name	Bariatric surgery for fertility‐sparing treatment of atypical hyperplasia and grade 1 cancer of the endometrium
Methods	RCT; parallel assignment; open‐label
Participants	36 women Age: 18 ‐ 41 years Inclusion criteria: 1. BMI ≥ 35 2. Diagnosis of grade 1 EEC or complex AEH 3. Clinical stage 1 disease ‐ no evidence of metastatic disease beyond the uterus by physical exam or imaging performed (MRI, CT) 4. ECOG status < 2 5. Desire for fertility conservation
Interventions	Bariatric surgery and progestin IUD
Outcomes	Primary outcome: recruitment rate (time frame: 3 years) Secondary outcome: completion of bariatric surgery; loss to follow‐up rate; completion of patient‐reported outcome questionnaires; complete response rate
Starting date	1 August 2020
Contact information	Gabrielle Ene gabrielle.ene@uhnresearch.ca
Notes	Estimated study completion date August 2023

NCT04046185.

Study name	Programmed death‐1 (PD‐1) inhibitor combined with progesterone treatment in endometrial cancer (ECCT)
Methods	RCT, parallel Assignment
Participants	60 participants Inclusion criteria: 1. Early EC patients (cancer confined in the endometrium, endometrioid histology, G1‐2). 2. Women want to spare fertility 3. Informed consent acquired 4. Age < 45 years
Interventions	Experimental: PD‐1 inhibitor and progesterone Toripalimab. 240 mg intravenous injection, every 3 weeks, 4 times plus MA tablets, 160 mg, orally, once a day Active comparator: progesterone MA tablets, 160 mg, orally, once a day
Outcomes	Primary outcomes: 1. Pathologic complete remission rate of endometrial curettage tissues (time frame: 6 months) 2. Pathologic partial remission rate of endometrial curettage tissues (time frame: 6 months) Secondary outcomes: 1. adverse effects (time frame: up to 1 year after treatment) 2. Pregnancy rate (time frame: up to 2 years after treatment)
Starting date	1 October 2019
Contact information	Xiaoping Wan, Dean, Professor, Shanghai First Maternity and Infant Hospital
Notes

NCT04362046.

Study name	Fertility sparing management of endometrial cancer and hyperplasia (FETCH)
Methods	Prospective; open‐label; single‐arm
Participants	30 women Age: 19‐39 years old Inclusion criteria Pathologist‐confirmed biopsy evidence of one of the following: Grade I EC or AEH MRI demonstrating less than 1/3 myometrial invasion if the patient has EC Desire to spare fertility
Interventions	Hysteroscopic uterine resection for patients who fail progestin therapy
Outcomes	Primary outcome: conception rate. Local disease control rate. Distant disease control rate Secondary outcome: adverse events
Starting date	May 2020
Contact information	Neeraj Mehra neeraj.mehra@ybc.ca
Notes	Estimated study completion date May 2028

NCT04491643.

Study name	Megestrol acetate plus rosuvastatin in young women with early endometrial carcinoma
Methods	Single‐group assignment; open‐label
Participants	43 women Age: 18 to 45 years Diagnosis based on hysteroscopy: histologically proved well‐differentiated EEC G1 without myometrial invasion
Interventions	MA + rosuvastatin Women will receive MA 160 mg and rosuvastatin 10 mg by mouth daily for at least 6 months Then every 3 months, hysteroscopy will be used to evaluate the endometrial condition.
Outcomes	Primary outcome: pathological response rate (time frame: 12 to 16 weeks) Secondary outcome Pathological response rate (time frame: 28 to 32 weeks) Pathological response duration (time frame: up to 2 years) Pathological response rate classified by different blood lipid level (time frame: up to 32 weeks) Toxicity evaluation (time frame: up to 32 weeks) Relapse rate (time frame: up to 2 years after the end of treatment ) Pregnancy rate (time frame: up to 2 years after the end of treatment)
Starting date	1 September 2020
Contact information	Xiaojun Chen; cxjlhjj@163.com Bingyi Yang; xiaomihaoku@163.com
Notes	Estimated primary completion date: 31 August 2023

NCT04491682.

Study name	Megestrol acetate plus rosuvastatin versus megestrol acetate in young women with atypical endometrial hyperplasia
Methods	RCT; parallel assignment
Participants	174 patients Age: 18‐45 years Confirmed pathological diagnosis based upon hysteroscopy
Interventions	Active comparator: MA alone: women will receive MA 160 mg by mouth daily for at least 6 months Experimental: MA + rosuvastatin; women will receive MA 160 mg and rosuvastatin 10 mg by mouth daily for at least 6 months
Outcomes	Primary outcome: pathological response rate (time frame: 12 to 16 weeks) Secondary outcomes Pathological response rate (Time Frame: 28 to 32 weeks) Pathological response duration (Time Frame: Up to 2 years)] Pathological response rate classified by different blood lipid level (Time Frame: Up to 32 weeks) Toxicity evaluation (Time Frame: Up to 32 weeks) Relapse rate (Time Frame: up to 2 years after the therapy for each patient ) Pregnancy rate (Time Frame: up to 2 years after the therapy for each patient)
Starting date	1 September 2020
Contact information	Xiaojun Chen; cxjlhjj@163.com Bingyi Yang; xiaomihaoku@163.com
Notes	Estimated primary completion date: 31 August 2023

NCT04607252.

Study name	Metformin plus megestrol acetate as a fertility‐sparing treatment in patients with atypical endometrial hyperplasia
Methods	RCT, parallel assignment, double masking
Participants	12 women Inclusion criteria 18‐45 years old; pathologically diagnosed with AEH for the first time; desire to preserve their fertility; no signs of suspicious myometrial invasion or extrauterine metastasis by enhanced MRI, enhanced CT or TVUS; no contraindication for metformin, MA or pregnancy; no hormone or metformin treatment within 6 months before entering the trial; not pregnant when participating in the trial; willing to follow the trial arrangement after being fully informed of all the risks and inconveniences caused by the trial
Interventions	Experimental: metformin (1500 mg per day) plus MA (160 mg per day) vs Active comparator: MA (160 mg per day).
Outcomes	Primary outcome: efficacy of the LNG‐IUD). Response rate at 1 year Secondary outcome: adverse events; quality‐of‐life score; molecular profile in pretreatment tissue to hysterectomy between responders and non‐responders to LNG‐IUD; long‐term survival; disease status; fertility outcomes
Starting date	First posted: 29 October 2020
Contact information	Xiaojun Chen, cxjlhjj@163.com
Notes

NCT04683237.

Study name	Liraglutide plus megestrol acetate in endometrial atypical hyperplasia
Methods	RCT, parallel assignment, open‐label
Participants	Women from 18 to 45 years Inclusion criteria BMI ≥ 28 kg/m2 Consent informed and signed Pathologically confirmed as AEH Have a strong desire to reproduce and ask for fertility preservation or those who insist on keeping the uterus despite no reproductive requirements Have good compliance and follow‐up conditions, and patients are willing to follow up in Obstetrics and Gynecology Hospital of Fudan University in time
Interventions	MA 160 mg/day orally) alone vs MA 160 mg/day orally + liraglutide(1.8 mg/d or the max tolerable dosage
Outcomes	Primary outcomes: pathological CR rates (time frame: from date of randomisation until the date of CR, assessed up to 28 weeks.). The 28‐week CR rates will be calculated in two groups Secondary outcomes: Weight loss (time frame: from date of randomisation until the date of CR, assessed up to 28 weeks). Weight changing in 28‐week's intervention, weight will be recorded every 4 weeks. Changing of insulin resistance (time frame: from date of randomisation until the date of CR, assessed up to 28 weeks.) Glycated hemoglobin, glucose tolerance test, insulin release test and C peptide release test will be performed at baseline and 28 weeks. During 12‐16 weeks, fasting glucose, insulin, glycated haemoglobin and fasting C peptide will be tested once. Improvement of chronic inflammation (time frame: from date of randomisation until the date of CR, assessed up to 28 weeks) improvement of chronic inflammation indications in 28 weeks. In detail, indicators including TNF‐α、IL‐1, IL‐6 and creatine kinase will be tested at 0, 12‐16th and 28th week. Time of pathological CR (time frame: from date of randomisation until the date of CR, assessed up to 2 years.) Time of histologic regression from AEH to proliferative or secretory endometrium Safety and side effects (time frame: from date of randomisation until the date of CR, assessed up to 2 years.) Adverse events related with liraglutide and MA. Number of participants with treatment‐related adverse events as assessed by Common Terminology for Adverse Events version 5.0 will be recorded, as well as incidence of adverse events. Relapse rates (time frame: up to 2 years after the treatment for each woman). All enrolled women will be followed up for 2 years. During the following‐up period, if patients recur after complete regression, they will be counted and the number of recurrence will be divided by number of patients followed up, then we can get the relapse rates. Comparison will be performed between two groups. Pregnancy outcomes (time frame: up to 2 years after the treatment for each patient). For patients who have a desire for fertility, pregnancies, births and related outcomes will be counted, and the rate of pregnancy will be counted as number of pregnancies/number of patients trying for fertility in the following period. Comparison will be performed between two groups
Starting date	20 March 2021. Estimated primary completion date: 30 November 2022
Contact information	Xiaojun Chen, cxjlhjj@163.com
Notes

NCT04897217.

Study name	Levonorgestrel‐releasing intrauterine system (LNG‐IUS) in the management of atypical endometrial hyperplasia
Methods	RCT, parallel assignment, open‐label
Participants	40 women Inclusion criteria Dilatation and curettage‐proven complex AEH only. Confirmed by pathology report. Normal renal function and liver function tests. Age 18 or older. The effects of MA on the developing human fetus at the recommended therapeutic dose are unknown. For this reason and because MA is known to be teratogenic, women of child‐bearing potential and men must agree to use adequate contraception (hormonal or barrier method of birth control; abstinence) prior to study entry and for the duration of study participation. Should a woman become pregnant or suspect she is pregnant while participating in this study, she should inform her treating physician immediately.
Interventions	Active comparator control arm: MA (160 mg by mouth daily) Experimental comparison arm: LNG‐IUD (with 52 mg progestin, releases 20 μ/daily)
Outcomes	Primary outcomes Histologic response scoring of biopsy specimens (time frame: 6 months). The primary outcome is a response score (0: no response, 1: incomplete response, 2: complete response): Score ‐1 will be assigned to specimens with progression of disease (ie. EC); Score 0 will be assigned to specimens with AEH; Score 1 will be assigned to specimens with benign hyperplasia and will be indicative partial response to therapy; Score 2 will be assigned to specimens without hyperplasia and will be indicative of CR to therapy. All study cases will be reviewed by a pathologist and all cases will be scored following the standard 2014 WHO classification of EH. Investigators will compare the estimated 90% confidence interval (CI) of the LNG‐IUS versus the standard therapy from the noninferiority trial to a predefined margin. Secondary outcomes Number of toxicities for each treatment (time frame: up to 8 months). Toxicities for each treatment modality based on standard toxicity scoring using NCI Common Terminology Criteria for Adverse Events (CTCAE) version 5.0. Investigators will compare the number of toxicities between the two treatment groups using Chi2 tests. Pathological response to treatment (time frame: up to 8 months). The pathological response variable will be treated as a binary variable, the Wilcoxon rank sum test will be used to explore the association between measurement of the endometrial stripe and the response outcome at each visit. Score ‐1 will be assigned to specimens with progression of disease (ie. EC); Score 0 will be assigned to specimens with AEH; Score 1 will be assigned to specimens with benign hyperplasia and will be indicative partial response to therapy; Score 2 will be assigned to specimens without hyperplasia and will be indicative of CR to therapy.
Starting date	April 2022
Contact information	Janelle Darby, Wake Forest Baptist Health Sciences
Notes	No fertility outcomes

NCT05172999.

Study name	Loxenatide plus LNG‐IUS in endometrial atypical hyperplasia
Methods	RCT, parallel assignment, open‐label
Participants	28 women from 18 to 45 years Inclusion criteria BMI ≥ 28kg/m2 Consent informed and signed Pathologically confirmed as AEH Have a strong desire to reproduce and ask for fertility preservation or those who insist on keeping the uterus despite no reproductive requirements Have good compliance and follow‐up conditions, and patients are willing to follow up in Obstetrics and Gynecology Hospital of Fudan University in time
Interventions	Active comparator: LNG‐IUS (for 6 months or longer till complete response) vs Experimental: LNG‐IUS + polyethylene glycol loxenatide Initiate injection of loxenatide will be 0.1 mg per week, if the patient can tolerate, the dose will be increased to 0.2 mg per week, or else 0.1 mg per week will be injected and the injection will last for no more than 28 weeks. If the patient cannot tolerate the least 0.1 mg/week, she must be excluded from this trial.
Outcomes	Primary outcomes Pathological complete response (CR) rates (time frame: from date of randomisation until the date of CR, assessed up to 16 weeks.) The 16‐week CR rates will be calculated in two groups Secondary outcomes 32‐week CR rates (time frame: from date of randomisation until the date of CR, assessed up to 32 weeks.) The 32‐week complete response rates will be calculated in two groups Weight change (time frame: from date of randomisation until the date of CR, assessed up to 32 weeks.) Weight change in 32‐week's intervention, weight will be recorded every month Changing of insulin resistance (time frame: from date of randomisation until the date of CR, assessed up to 32 weeks.) Indicators of insulin resistance will be assessed at baseline and every 12‐16 weeks. Change of chronic inflammation status (time frame: from date of randomisation until the date of CR, assessed up to 32 weeks.) Change of chronic inflammation indications (tumour necrosis factor (TNF‐α), interleukin‐1(IL‐1), interleukin‐6 (IL‐6)) in 32 weeks Change of ovarian functions (time frame: from date of randomisation until 6 months after treatment is over) AMH (anti‐mullerian hormone ) will be assessed at baseline and every 12‐16 weeks Time of pathological CR (time frame: from date of randomisation until the date of CR, assessed up to 2 years). Time of histologic regression from AEH to proliferative or secretory endometrium Safety and side effects (time frame: from date of randomisation until the date of CR, assessed up to 2 years). Adverse events related with liraglutide and LNG‐IUS. Number of participants with treatment‐related adverse events as assessed by CTCAE v5.0 will be recorded, as well as incidence of adverse events. Relapse rates (time frame: up to 2 years after the treatment for each patient). All enrolled patients will be followed up for 2 years. During the following‐up period, if patients recur after complete regression, they will be counted and the number of recurrences will be divided by number of patients followed up, then we can get the relapse rates. Comparison will be performed between two groups. Pregnancy outcomes (time frame: up to 2 years after the treatment for each patient). For patients have a desire for fertility, pregnancies, births and related outcomes will be counted, and the rate of pregnancy will be counted as number of pregnancies/number of patients trying for fertility in the following period. Comparison will be performed between two groups
Starting date	8 January 2022
Contact information	Xiaojun Chen, cxjlhjj@163.com
Notes

NCT05247268.

Study name	Gonadotropin‐releasing hormone agonist (GnRHa) plus letrozole in young women with early endometrial cancer
Methods	This will be a multicenter RCT to evaluate the treatment effects and adverse events of GnRHa plus AIs compared with MA/MPA in primary EEC patients.In this study, young patients (18‐45 years) diagnosed as EEC for the first time seeking for fertility preserving treatment at the Obstetrics and Gynecology Hospital of Fudan University were screened. Patients were randomly assigned (1:1) to GnRHa+letrozole group (triprorelin acetate, intramuscular injection of 3.75 mg was given 4 weeks apart and the maximum use are 6 courses. Letrozole, 2.5 mg oral daily and no more than 24 weeks) or MA/MPA group (160 mg oral MA daily or 500 mg oral MPA daily)
Participants	Inclusion criteria Consent informed and signed Primarily have a confirmed diagnosis of early‐stage EC (endometrioid, grade I, without myometrial invasion) ) based upon endometrial biopsy, diagnostic curettage or hysteroscopy No signs of suspicious myometrial invasion or extrauterine metastasis by enhanced MRI, enhanced CT or TVUS Have a strong desire for remaining reproductive function or uterus Have good compliance and follow‐up conditions, and patients are willing to follow up in Obstetrics and Gynecology Hospital of Fudan University in time Exclusion criteria Pathologically confirmed as EC with suspicious myometrial invasion or extrauterine metastasis Recurrent EC Combined with severe medical disease or severely impaired liver and kidney function Patients with other types of EC or other malignant tumours of the reproductive system; patients with breast cancer or other hormone‐ dependent tumours that cannot be used with progesterone Those who require hysterectomy or other methods other than conservative treatment with drugs Known or suspected pregnancy Contraindication for MPA, MA, triprorelin acetate, letrozole or pregnancy Hormone treatment within 3 months before entering the trial; Acute severe disease such as stroke or heart infarction or a history of thrombosis disease Smoker (> 15 cigarettes a day)
Interventions	Intervention: GnRHa + letrozole group (triprorelin acetate, intramuscular injection of 3.75 mg was given 4 weeks apart and the maximum use are 6 courses. Letrozole, 2.5 mg oral daily and no more than 24 weeks) Control: MA/MPA group (160 mg oral MA daily or 500 mg oral MPA daily)
Outcomes	CR rates within 16 weeks of treatment
Starting date	11 March 2022
Contact information	Xiaojun Chen
Notes

NCT05492487.

Study name	A pilot study on fertility conservative treatment of AEH in Singapore
Methods	Participants will be randomised into either the megace group or the Mirena group. There will be sealed envelopes containing a paper with either Mirena or megace printed on each of them. Participants can chose an envelope at random. The patient and PI will not be blinded of the selection in the study.
Participants	Inclusion criteria All women diagnosed with AH aged 21 years to 40 years Keen for fertility‐preserving treatment Exclusion criteria Patients who are currently undergoing treatment for AH Patients with a history of EC
Interventions	Intervention: megace use Control: Mirena use
Outcomes	Rate of regression. Time taken for regression of the disease
Starting date	3 January 2020
Contact information	Jessie Phoon
Notes

NCT05565573.

Study name	Medroxyprogesterone acetate vs LNG‐IUS in early‐stage endometrioid carcinoma and atypical hyperplasia patients
Methods	The present study is a RCT. It aims to compared the therapeutic effect of MPA and LNG‐IUS in early‐stage EC and AH patients.
Participants	Inclusion criteria Age ≤ 45 years old; Diagnostic curettage pathology is highly differentiated EAC; pelvic augmentation MRI or/and vaginal colour ultrasound Doppler examination, tumour confined endometrium; No suspicious metastatic lesions; Endometrial pathological tissue specimens with strong positive expression of oestrogen and progesterone receptors; Blood CA125 is normal; No progesterone therapy or contraindications to pregnancy; The patient has a strong desire to preserve fertility, and the patient should fully understand that fertility preservation treatment is not the standard treatment method; Conditional acceptance of close follow‐up. Exclusion criteria Type II EC; Patients with abnormal liver and kidney function and other contraindications to progesterone therapy; Other parts of the body are concurrent or successively complicated by other malignant tumours; The patient and his/her family are unable to sign the informed consent form for any reason; Unconditional follow‐up
Interventions	Experimental: MPA. Administered MPA at a dosage of 500 mg/d concurrently Control: go through LNG‐IUS insertion
Outcomes	CR rate
Starting date	1 November 2022
Contact information	Hengzi Sun
Notes

Zhao 2022.

Study name	Levonorgestrel‐releasing intrauterine device plus metformin, or megestrol acetate plus metformin for fertility‐sparing treatment of AEH and early endometrial carcinoma: a prospectiva, randomized, blind‐endpoint design
Methods	This will be an open‐label, 2‐armed, randomised, phase‐II single‐center trial of LNG‐IUD plus metformin or MA plus metformin. A total of 88 participants will be randomly assigned into 2 treatment arms in a 1:1 ratio. Clinical, laboratory, ultrasound and radiology data, will be collected at baseline, and then at 3, 6, 9, 12, 18, and 24 months. EC biomarkers will be collected at baseline. The primary aim is to determine the efficacy of a LNG‐IUD plus metformin, or MA plus metformin in achieving pathological CR at 12 months, as well as post‐treatment pregnancy outcomes and recurrence rate. The secondary aims are to predict the response to an LNG‐IUD plus metformin and MA plus metformin via clinical, blood, and tissue predictive biomarkers
Participants	Inclusion criteria Women with a BMI > 24 kg/m2 who wish to maintain fertility Aged over 18 years and under 45 years at the time of randomisation No contraindications to pregnancy Histologically confirmed complex EHA or grade 1 EC on a curette or endometrial biopsy tissue specimen CT of the pelvis, abdomen, and chest (or chest X‐ray) indicates the absence of extrauterine disease MRI shows less than 50% of myometrial invasion (women with EC only). No evidence of lymph vascular invasion on curetting specimens. FIGO stage IA EC Serum CA 125 level ≤ 30 U/mL No hypersensitivity or contraindications to LNG‐IUD, MA, or metformin Ability to comply with endometrial biopsies at specified intervals No use of metformin for at least 2 years before study participation No history of LNG‐IUD use No prior use of medroxyprogesterone acetate The primary physician recommended treatment with LNG‐IU or MA for EHA or grade 1 EC. A negative pregnancy test within 7 days of starting treatment Understanding the study design, risks and benefits, and providing informed consent Exclusion criteria Patients who meet any of the exclusion criteria at baseline will be excluded. Eastern Cooperative Oncology Group (ECOG) performance status > 3 MRI shows grade 1 EC with myometrial invasion > 50%, grade 2 or grade 3 EC Histological cell type other than EC (sarcoma or high‐risk cell type, e.g., papillary serous or clear cell) Prior treatment for EC or EHA A history of pelvic or abdominal radiotherapy Renal impairment (creatinine > 150 µmol/L [1.7 mg/dL]), acute pulmonary oedema, or liver failure Unwilling to have additional endometrial biopsies, or unable to attend 3‐monthly clinical assessments Unable to provide informed consent or complete questionnaires Medical imaging shows evidence of extrauterine spread A congenital or acquired uterine anomaly which distorts the uterine cavity Acute pelvic inflammatory disease Concurrent with different cancer Chronic (daily use for > 1 month) use of cimetidine (significant increase in metformin concentration and risk of lactic acidosis) Use of Iodinated contrast agents within the past 48 h (significant increase in metformin concentration and risk of lactic acidosis) Pregnant or lactating History of hypersensitivity to metformin or discontinuation due to adverse effects High risk of thrombosis Surgery within a week before starting treatment Current or history of cerebral and/or cardiac infarction or thrombophlebitis Valvular heart disease, atrial fibrillation, endocarditis, or other serious cardiac conditions Currently receiving hormone treatments
Interventions	Intervention:LNG‐IUD plus metformin Control: MA plus metformin in achieving pathological CR
Outcomes	The primary aim is to determine the efficacy of a LNG‐IUD plus metformin, or MA plus metformin in achieving pathological CR at 12 months, as well as post‐treatment pregnancy outcomes and recurrence rate. The secondary aims are to predict the response to an LNG‐IUD plus metformin and MA plus metformin via clinical, blood, and tissue predictive biomarkers.
Starting date	1 January 2020
Contact information	Jumin Niu, Email: moc.361@700xyyzgt.
Notes

A(E)H: atypical (endometrial) hyperplasia; BMI: Body mass index; CR: complete (pathological) response/remission; CT: computed tomography; DFS: disease‐free survival; EAC: endometriod adenocarcinoma; ECOG: Eastern Cooperative Oncology Group; (E)EC: (endometroid) endometrial cancer; GnRH‐a: gonadotropin‐releasing hormone agonist; LNG‐IUD/S: levonorgestrel intrauterine device/system; MA: megestrol acetate; MPA: medroxyprogesterone acetate; MRI: magnetic resonance imaging; RCT: randomised controlled trial; RFS: recurrence‐free survival; TVUS: transvaginal ultrasound

Differences between protocol and review

Our protocol is published in the Cochrane Library (Fernandez‐Montoli 2018).

Search strategy

The protocol specified that conference proceedings and abstracts would be searched through ZETOC (http://zetoc.jisc.ac.uk/), but it was not possible to access this payment resource, neither through the authors’ institutions nor through Cochrane Gynaecological Oncology, Neuro‐oncology and Orphan Cancer. We could not search through Worldcat (www.worldcat.org/advancedsearch).

Objectives

In the protocol, the objective was: "To compare the effectiveness and safety of fertility‐sparing treatments, including pharmacological interventions (e.g. oral progestin, levonorgestrel IUS, metformin) and bariatric surgery." In the review, we added 'hysteroscopic surgery', as it had great clinical relevance for fertility‐sparing treatment.

Types of participants

Although the protocol stated as inclusion criteria "Women 18 years of age or older, who desire fertility conservation", we have included some studies with adult women of conceiving age (mean age of participants < 50 years) treated with conservative management regardless of fertility desire, as we considered that these studies’ results could be applicable to the review question. We excluded studies with the mean age of women older than 50 years, and studies that only included women with co‐morbidities, or not candidates for surgical treatment.

Types of studies included

In our protocol, we planned to include non‐randomised studies (NRSs), in the case that we did not find RCTs. However, the searches did identify some RCTs, although only a few. For this reason, we described NRSs narratively in the review, and the risk of bias of these NRSs is not shown in the text.

In our protocol we did not state that studies with cointerventions would be included, as long as cointerventions were balanced in all study arms. We added this criterion in the Methods section of this review.

In the protocol we did not specify that we considered it appropriate to pull together data from participants with atypical endometrial hyperplasia (AEH; pre‐malignant lesion) and stage IA1 endometrial cancer in the same analysis. We added this criterion in the Methods section of the review.

In the Methods section (Types of interventions) we added 'hysteroscopic surgery' to the list of surgical techniques, as it is a relevant technique for this review.

Management of database records

In the protocol, we did not outline any plan for the management of database records. However, as Covidence is available for Cochrane members, we imported and screened all database records in Covidence.

Contacting study authors

We contacted the corresponding authors of seven studies to obtain or clarify data related to their research. In one case, the authors informed us that they were unable to share the results at that time. However, in two instances involving included studies, the authors were able to provide the requested information (Shan 2014; Yang 2020). We received feedback regarding the outcomes of live birth rate, pregnancy rate, and surgery for persistent/progressive disease (hysterectomies). However, we only obtained information on adverse events from Yang 2020.

Unfortunately, we did not receive a response from the remaining study authors.

Rating overall risk of bias

In the protocol, we did not specify a rating of overall risk of bias. We planned to judge the risk of bias of the domains in each RCT according to the Cochrane tool for assessing risk of bias (RoB 1; Higgins 2017). However, in the review, we rated the overall bias judgment of included RCTs (Higgins 2017).

If all the domains were labelled as a low‐risk study, a RCT was labelled as low risk. An RCT at high risk in one or more domains was labelled as a high‐risk study; the trial was judged to have unclear risk of bias if it was labelled as unclear risk in one or more domains (Higgins 2017).

Risk of bias in non‐randomised studies

In our protocol, we planned to include NRSs, in case we did not find any RCTs. We planned to assess the risk of bias of NRSs, using the ROBINS I tool (Sterne 2016). However, we found some RCTs. For this reason, NRSs are described narratively in the review, and risk of bias for NRSs is not shown in the text.

Sensitivity analysis

In the protocol, we planned to perform a sensitivity analysis for the primary outcome if there were sufficient studies included that took into account the risk of bias, the source of the data (type of studies included) and the decisions taken according to the heterogeneity analysis. We were unable to conduct a sensitivity analysis because only two studies were included in each comparison.

Subgroup analysis

In the protocol, we planned to perform a subgroup analysis, grouping the trials according to histology: treatment of endometrial cancer/AEH. However, we could not perform a subgroup analysis because we only found information on AEH in the RCTs included. So, we could not compare AEH with endometrial cancer cases.

Comparisons included in the summary of findings tables

Our protocol did not specify the comparisons included in the summary of findings tables. After involving the review’s co‐authors and relevant stakeholders, we decided which comparisons were the most clinically relevant. We included summary of findings tables for the following comparisons.

• Metformin + progestin compared to progestin

• Levonorgestrel intrauterine system compared to oral progestin

• Oral progestin plus levonorgestrel intrauterine system (IUS) compared to oral progestin

Summary of findings' tables were not to be produced if no studies were found in a specific comparison. This fact does not mean that a comparison is not clinically relevant.

Contributions of authors

Maria‐ Eulalia Fernández‐ Montoli conceptualised the overall review project.

MEF, Nayanar Contreras, Paula Verdaguer, Judith Lleberia and Carla Julià conducted the investigation.

NC, PV, JL and CJ searched for 'grey literature'.

MEF, NC, PV, JL and CJ independently assessed the eligibility of retrieved papers and resolved any disagreements by discussion.

Two review authors (MEF or Jordi Sabadell) and (NC, PV, JL or CJ) independently collected data in an extraction form for the review. We consulted another review author (JS) if necessary.

MEF and JS did the formal analysis by assessing the risk of bias of the included studies, as well as data analysis and interpretation.

MEF and JS wrote the review; original draft preparation and review editing

Sources of support

Internal sources

• No sources of support, Other

  No sources of support

External sources

• No sources of support, Other

  No sources of support

Declarations of interest

Maria‐Eulalia Fernandez‐Montoli: none known
Jordi Sabadell: none known
Nayanar‐Adela Contreras Perez: none known
Paula Verdaguer: none known
Carla Julià Torres: none know
Judith Lleberia; none known

New