Levonorgestrel-releasing intrauterine device therapy vs oral progestin treatment for reproductive-aged patients with endometrial intraepithelial neoplasia: a systematic review and meta-analysis

Abstract

Background

We conducted a systematic review and meta-analysis to examine outcomes of patients with endometrial intraepithelial neoplasia treated with oral progestins or a levonorgestrel-releasing intrauterine device (IUD).

Methods

We conducted a systematic review across 5 databases to examine outcomes of progestational treatment (oral progestins or levonorgestrel-releasing IUD) for patients with endometrial intraepithelial neoplasia. The primary outcome was the best complete response rate within 12 months of primary progestational treatment. Sensitivity analyses were performed by removing studies with extreme effect sizes. Secondary outcomes included the pooled pregnancy rate.

Results

We identified 21 eligible studies, including 824 premenopausal patients with endometrial intraepithelial neoplasia, for our meta-analysis. Among these, 459 patients received oral progestin, and 365 patients received levonorgestrel-releasing IUD as a primary progestational treatment. The pooled best complete response proportion within 12 months was 82% (95% confidence interval [CI] = 69% to 91%) following oral progestin treatment and 95% (95% CI = 81% to 99%) following levonorgestrel-releasing IUD treatment. After removing outlier studies, the pooled proportion was 86% (95% CI = 75% to 92%) for the oral progestin group and 96% (95% CI = 91% to 99%) for the levonorgestrel-releasing IUD group, with reduced heterogeneity. The pooled pregnancy rate was 50% (95% CI = 35% to 65%) after oral progestin and 35% (95% CI = 23% to 49%) after levonorgestrel-releasing IUD treatment.

Conclusions

This meta-analysis provides data on the effectiveness of oral progestins and levonorgestrel-releasing IUD treatment within 12 months of treatment among premenopausal patients with endometrial intraepithelial neoplasia. Although based on small numbers, the rate of pregnancy after treatment is modest. These data may be beneficial for selecting progestational therapies that allow fertility preservation for patients with endometrial intraepithelial neoplasia.

Endometrial intraepithelial neoplasia, formerly atypical endometrial hyperplasia, is a precursor to endometrial cancer, which is the most common gynecologic cancer in the United States and often occurs in postmenopausal women. Although only 12% of incident cases of endometrial cancer in the United States occur in women aged younger than 50 years (1), the incidence of early stage, low-grade endometrial cancer is increasing among women in their 30s (2), and recent trends toward delayed childbearing predict a growing population of premenopausal patients with endometrial intraepithelial neoplasia who may desire fertility preservation (3). Importantly, among women with endometrial intraepithelial neoplasia, 43% have occult cancer at the time of diagnosis, and 30% of untreated patients progress to cancer (4,5).

Endometrial intraepithelial neoplasia localizes within the endometrium thus allowing for consideration of fertility-sparing treatments for patients who desire future childbearing. As an alternative to conventional surgical management, progestational therapies are an option for conservative therapy (6-8). The 2 prominent types of progestational therapies include local therapy with the levonorgestrel-releasing intrauterine device (IUD) or systemic therapy with oral progestins (7,8).

There is a dearth of research evaluating the efficacy of different routes of progestational therapy (9,10). Specifically, only 1 randomized controlled trial (RCT) has directly compared the efficacy of oral progestins vs levonorgestrel-releasing IUD in treating premenopausal patients with endometrial intraepithelial neoplasia (8,9,11). However, the trial was limited by a small sample size with only 12% (n = 19) of the participants being pathologically confirmed to have endometrial intraepithelial neoplasia (11). Much of the available data comparing oral and intrauterine progestin therapy is based on observational studies and case series (12). Further, prior meta-analyses have a number of limitations such as inclusion of cases of endometrial hyperplasia without atypia or endometrial cancer (10,12-14), failing to distinguish premenopausal and postmenopausal patients, and inclusion of a wide range of assessment times, making it difficult to compare the effects of treatment (10,12,14,15).

Given the increased use of fertility-sparing treatments in premenopausal patients with endometrial intraepithelial neoplasia over the past decade (3), and the paucity of high-quality evidence on the efficacy of oral progestins and levonorgestrel-releasing IUD in these patients, we conducted a systematic review and meta-analysis to synthesize evidence in this area. We examined the complete response rate and other pertinent secondary outcomes following 12 months of treatment with either oral or intrauterine progestational therapy in premenopausal patients with endometrial intraepithelial neoplasia.

Methods

This meta-analysis was conducted using the Preferred Reporting Items for Systematic Reviews and Meta-Analysis reporting guidelines (16) and registered in the International Prospective Register of Systematic Reviews under the registration number CRD42022379996.

Search strategy

Literature searches to inform the meta-analysis were conducted by a Columbia University Health Sciences Library Informationist (JU) on November 28, 2022. Five databases were searched for relevant literature: PubMed, Embase, Cochrane Library, Scopus, and Web of Science. The strategy comprised of a combination of search strings related to IUD and/or progestin and endometrial hyperplasia. No filters were applied to the search results. In addition to the databases, a search of ClinicalTrials.gov was performed to identify ongoing or recently completed clinical trials. The full search strategies can be found in Supplementary Material 1 (available online).

All records found via the search were exported to the citation manager EndNote (version X9). Duplicates were identified and removed in the EndNote library, and the deduplicated library was then imported into Covidence (17) to facilitate the process of screening, appraising, and extracting information from relevant articles for the meta-analysis.

Study selection

We included 1) endometrial intraepithelial neoplasia patients aged younger than 50 years per the World Health Organization’s age cutoff for reproductive age; 2) patients with pathologically confirmed endometrial intraepithelial neoplasia; 3) studies examining conservative, nonsurgical treatment of endometrial intraepithelial neoplasia patients with 1 or more treatment arms that included progestin therapy; 4) studies in any country; 5) any study design but with at least 10 patients of endometrial intraepithelial neoplasia in a treatment group; 6) studies that included the best complete response proportion within 1 year among those who used oral progestins or who used levonorgestrel-releasing IUD; and 7) full text papers in English.

The primary exposure was the type of progestational treatment (oral progestins or levonorgestrel-releasing IUD). The primary outcome was the pooled proportion of the best complete response within 1 year of primary progestational treatment. Secondary outcomes included complete response at 6 months from the primary hormonal treatment, number of pregnancies and live births, number of hysterectomies, number of patients who progressed to cancer, and number of recurrences among patients who achieved an initial complete response. In addition, we examined well-described, serious side effects associated with progestational therapy, including venous thromboembolism.

Data extraction and quality assessment

Two reviewers (YS and JSF) independently performed the study review and selection, data extraction, and quality assessment. Discrepancies were resolved via discussion and/or consultation with a third reviewer. Extracted data included author’s name; year published; country; type of progestin therapy; hormonal regimen used; adjunct therapy use such as metformin, gonadotropin-releasing hormone (GnRH) agonist, or transcervical resection; the best complete response within 1 year of primary hormonal treatment; complete response at 6 months from the primary hormonal treatment; pregnancies among patients who attempted to conceive; live births among patients who attempted to conceive; hysterectomy; recurrence after complete regression; and venous thromboembolic events. Transcervical resection was defined as a method of using hysteroscopy. We did not consider the assessment time frame for the secondary outcomes except for complete response at 6 months.

We assessed the risk of bias in each study using Joanna Briggs Institute (JBI) critical appraisal checklist for cohort studies and for RCTs (18). The JBI critical appraisal checklist for cohort studies includes 11 items, with a maximum overall score of 11 assessing the methodological quality of a study and determining the extent to which a study has addressed the possibility of bias in its design. Similar domains were assessed in the JBI critical appraisal checklist for RCT, which is composed of 13 items, with a maximum overall score of 13 (Supplemental Tables 1 and 2, available online). Higher scores indicate better study quality.

Statistical analysis

Outcome proportions were calculated by dividing the number of patients who had the outcome of interest following hormonal treatment divided by the total number of patients who received the treatment. For individual studies, 95% confidence intervals (CIs) were calculated with the Clopper–Pearson exact method (19). This method accommodates small samples and extreme proportions (close to 0% or 100%).

Logit transformation was used for proportions before the meta-analysis was performed. Then, generalized logistic mixed-effect models were developed to estimate the pooled effect (19). We used the metaprop function in R to estimate the pooled proportion of each primary and secondary outcome. Because we anticipated considerable heterogeneity in the included studies, we used Hartung–Knapp adjustments for the random-effects model to provide more conservative estimates of the 95% confidence intervals around the pooled proportions (19-21). We used I² statistics to assess interstudy heterogeneity.

As an additional outcome, the pooled odds ratio (OR) was calculated by comparing the best complete response within 12 months following levonorgestrel-releasing IUD with oral progestin, including only 3 studies that had both treatment arms in the same study. We used the Mantel–Haensel method for pooling and Hartung–Knapp adjustments for the random-effects model to calculate the 95% confidence interval around the pooled effect using the metabin function in R (19).

Sensitivity analyses were conducted to examine the robustness of our findings. These sensitivity analyses examined the complete response proportion within 12 months and secondary outcomes (complete response rate at 6 months and the pregnancy rate) by removing outlier studies that could bias the results. The outlier studies were those that contributed the most to the heterogeneity of meta-analysis results as identified by using the leave-one-out method and alternatively by using a Baujat plot (19).

A subgroup analysis comparing the pooled proportion of the best complete response within 12 months by GnRH agonist use among levonorgestrel-releasing IUD treatment groups was conducted. Sensitivity analysis was also performed by removing 1 outlier study.

We performed a publication bias test using a funnel plot and the Egger test (19,22). For the levonorgestrel-releasing IUD group, given that the number of studies was small, Peter test to detect bias was applied (19,23). A 2-sided P value less than .05 was considered statistically significant. All statistical analyses were performed using R software, version 4.2.2 (R Core Team, Vienna, Austria).

Results

Search results and characteristics

A total of 8524 records were found via the search of the 5 databases and ClinicalTrials.gov. After removing 4063 duplicates, we screened the remaining 4461 unique records for inclusion in the meta-analysis. After title and abstract screening and full-text review, 21 studies were ultimately included (Figure 1); 20 studies were classified as cohort studies (24-43), and 1 study was an RCT (44) (Table 1). Among the cohort studies, 15 studies were single exposure cohort studies (45). The studies were reported from 7 countries, including 3 from Europe (France, Italy, Russia), 3 from Asia (China, Iran, Japan), and 1 from Australia. There were 16 oral progestin arms and 8 levonorgestrel-releasing IUD arms from all the studies included. Use of adjunctive therapy is listed in Table 1. We ultimately included 824 premenopausal patients with endometrial intraepithelial neoplasia of whom 459 patients received oral progestin and 365 patients received levonorgestrel-releasing IUD as their primary progestational treatment.

Figure 1.

Preferred Reporting Items for Systemic Reviews and Meta-Analysis flow diagram.

Table 1.

Characteristics of studies included in final meta-analysis

						Outcomes, No. of patients (%)
Study	Year	Country	Progestin therapy group	Regimen	Adjunct therapy (Met/GnRH/TCR)	Best complete regression within 12 mo	Complete regression at 6 mo	Pregnancya	Live birth	Hysterectomy	Recurrence after CR	VTE
Cohort studies (n = 20)
Chen et al. (24)	2022	China	Oral	MA, 160 mg/d or MPA, 500 mg/d	None	16/19 (84.2)	NR	3/9 (33.3)	2/9 (22.2)	NR	3/16 (18.7)	0/19 (0.0)
			IUD	IUD	GnRH agonist	20/21 (95.2)	NR	4/14 (28.6)	2/14 (14.2)	NR	2/20 (10.0)	0/21 (0.0)
Jing et al. (25)d	2022	China	Oral	MA, 160 mg/d	Met	31/31 (100.0)	28/31 (90.3)	NR	14/31 (45.2)	4/11 (36.4)	11/31 (35.5)	0/31 (0.0)
Akhavan et al. (26)d	2021	Iran	Oral	Megestrol, 160 mg	None	9/24 (37.5)	6/38 (15.8)	8/25 (32.0)	7/25 (28.0)	15/50 (30.0)	NR	NR
Fang et al. (27)	2021	China	IUD	IUD	None	11/20 (55.0)	11/20 (55.0)	5/11 (45.5)	2/11 (18.2)	NR	NR	NR
Novikova et al. (28)	2021	Russia	IUD	IUD	GnRH agonist	19/20 (95.0)	NR	NR	NR	NR	NR	NR
			IUD	IUD	None	166/169 (98.2)	NR	NR	NR	NR	NR	NR
			Oral	MPA, 500 mg/d	None	34/39 (87.0)	NR	NR	NR	NR	NR	NR
Raffone et al. (29)d	2021	Italy	IUD	IUD	TCR	37/37 (100.0)	37/37 (100.0)	10/37 (27.0)	NR	NR	NR	NR
Casadio et al. (30)d	2020	Italy	Oral	MA, 160 mg/d	TCR	46/46 (100.0)	NR	31/45 (68.9)	21/45 (46.7)	4/46 (8.7)	4/46 (8.7)	NR
Giampaolino et al. (31)d	2019	Italy	IUD	IUD	TCR	51/55 (92.7)	51/55 (92.7)	10/25 (40.0)	10/25 (40.0)	1/47 (2.1)	2/47 (4.3)	NR
Tamauchi et al. (32)d	2018	Japan	Oral	MPA, 600 mg/d	None	22/30 (78.6)	22/30 (78.6)	9/15 (60.0)	7/15 (46.7)	NR	14/28 (50.0)	NR
Chen et al. (33)d	2016	China	Oral	MA, 160-480mg/d or MPA, 250-500mg/d	GnRH agonist (depends on the response)	7/16 (43.8)	7/16 (43.8)	9/16 (56.3)	6/16 (37.5)	4/16 (25.0)	3/16 (18.8)	NR
Mitsuhashi et al. (34)d	2016	Japan	Oral	MPA, 400 mg/d	Met	16/17 (94.1)	NR	NR	NR	NR	0/16 (0.0)	0/17 (0.0)
Ohyagi-Hara et al. (35)d	2015	Japan	Oral	MPA, 400-600mg/d	None	9/11 (81.8)	NR	4/NA (NA)b	4/NA (NA)c	NR	0/11 (0.0)	NR
Brownfoot et al. (36)	2014	Australia	Oral	MPA	None	24/32 (75.0)	NR	NR	NR	NR	NR	NR
			IUD	IUD	None	8/10 (80.0)	NR	NR	NR	NR	NR	NR
Shan et al. (37)	2014	China	Oral	MA, 160 mg/d	Met	8/16 (50.0)	NR	NR	NR	NR	NR	NR
Koskas et al. (38)d	2012	France	Oral	MPA, MA, CA, NoA, or lynestrenol	None	12/14 (85.7)	12/14 (85.7)	6/14 (42.9)	5/14 (35.7)	4/14 (28.6)	1/12 (8.3)	NR
Pashov et al. (39)d	2012	Russia	IUD	IUD	GnRH agonist	13/13 (100.0)c	NR	NR	NR	NR	NR	NR
Minig et al. (40)d	2011	Italy	IUD	IUD	GnRH agonist	19/20 (95.0)	18/20 (90.0)	8/19 (42.1)	6/19 (31.6)	NR	4/19 (21.1)	0/20 (0.0)
Yu et al. (41)d	2009	China	Oral	MPA, 100-500mg/d	None	14/17 (82.4)	NR	4/10 (40.0)	3/10 (30.0)	NR	3/14 (21.4)	NR
Ushijima et al. (42)d	2007	Japan	Oral	MPA, 600 mg/d	None	16/17 (94.1)	14/17 (82.4)	7/NA	4/NA (NA) b	NR	6/16 (37.5)	0/17 (0.0)
Kaku et al. (43)d	2001	Japan	Oral	MPA, 100-800mg/d	None	12/18 (66.7)	12/18 (66.7)	5/NA	4/NA (NA) b	7/18 (38.9)	2/13 (15.4)	0/18 (0.0)
Randomized controlled trials (n = 1)
Yang et al. (44)	2020	China	Oral	MA, 160 mg/d	Met	44/55 (72.7)	NR	NR	NR	NR	NR	NR
			Oral	MA, 160 mg/d	None	39/57 (68.4)	NR	NR	NR	NR	NR	NR

a

The denominator is the total number of patients who attempted to conceive. CA = chlormadinone acetate; CR = complete regression; GnRH = gonadotropin-releasing hormone; IUD = levonorgestrel intrauterine device; MA = megestrol acetate; Met = metformin; MPA = medroxyprogesterone acetate; NA = not applicable; NoA = nomegestrol acetate; NR = not reported; Oral = oral progestin; TCR = transcervical resection; VTE = venous thromboembolism.

b

There is no information on the number of patients who attempted to conceive.

c

In this study, the assessment was performed at 12-14 months.

d

A single exposure cohort study.

Primary outcome

The pooled best complete response proportion within 12 months of oral progestin treatment was 0.82 (95% CI = 0.69 to 0.91) (Figure 2); however, there was statistically significant between-study heterogeneity (I² = 58%, P < .01). The pooled best complete response proportion within 12 months for levonorgestrel-releasing treatment was 0.95 (95% CI = 0.81 to 0.99), and similarly, there was significant between-study heterogeneity (I² = 79%, P < .01) (Figure 2).

Figure 2.

Forest plots showing the best complete response proportion within 12 months. CI = confidence interval; LNG-IUD = levonorgestrel intrauterine device.

Secondary outcomes

The pooled complete response proportion at 6 months after oral progestin treatment was 0.68 (95% CI = 0.38 to 0.88) compared with 0.92 (95% CI = 0.39 to 0.99) after levonorgestrel-releasing IUD treatment (Figure 3).

Figure 3.

Forest plots showing the best complete response proportion at 6 months. CI = confidence interval; LNG-IUD = Levonorgestrel intrauterine device.

Figure 4, A and B, show the pregnancy and live birth outcomes, respectively, where the denominator of the proportion is the number of patients with a complete response who attempted to conceive. The pooled pregnancy proportion following oral progestin treatment was 0.50 (95% CI = 0.35 to 0.65), and the proportion following levonorgestrel-releasing IUD treatment was 0.35 (95% CI = 0.23 to 0.49) (Figure 4, A). The pooled live birth proportion following oral progestin treatment was 0.39 (95% CI = 0.31 to 0.49), while the proportion following levonorgestrel-releasing IUD treatment was 0.29 (95% CI = 0.15 to 0.49) (Figure 4, B).

Figure 4.

Forest plot showing secondary outcomes. A) Pregnancy. B) Live birth. C) Hysterectomy. D) Progressed to cancer. E) Recurrence after complete response. F) Venous thromboembolism. CI = confidence interval; CR = complete response; LNG-IUD = levonorgestrel intrauterine device; VTE = venous thromboembolism.

The pooled hysterectomy, progression to cancer, and recurrence after complete response proportions following oral progestin treatment were 0.22 (95% CI = 0.12 to 0.36), 0.10 (95% CI = 0.01 to 0.43), and 0.16 (95% CI = 0.08 to 0.28), respectively (Figure 4, C-E). Only 1 study was identified for each of these outcomes with levonorgestrel-releasing IUD treatment; thus, we were unable to calculate a pooled effect size. There were no patients of venous thromboembolism in the included studies (Figure 4, F).

For studies that examined both levonorgestrel-releasing IUD and oral progestin, a pooled odds ratio of the best response within 12 months of treatment was estimated. There was no statistically significant association between type of treatment (levonorgestrel-releasing IUD vs oral progestin) and complete response within 12 months of hormonal treatment (OR = 3.66, 95% CI = 0.41 to 32.64) (Supplementary Figure 1, available online).

Sensitivity and additional analyses

Sensitivity analyses were performed after removing outlier studies with extreme effect sizes. The pooled estimate of the best complete response within 12 months of oral progestin treatment was 0.84 (95% CI = 0.72 to 0.92) with lower heterogeneity (Table 2) when omitting Akhavan et al. (26) by the leave-one-out method (sensitivity analysis 1) (Table 2; Supplementary Figure 2, A, available online). Using a Baujat plot (sensitivity analysis 2), 2 outliers were identified and removed resulting in a pooled proportion of 0.86 (95% CI = 0.75 to 0.92) with an I² of 25% (Table 2; Supplementary Figures 3, A, and 4, A, available online). For levonorgestrel-releasing IUD treatment, the pooled proportion of the best complete response within 12 months was 0.96 (95% CI = 0.91 to 0.99) with lower heterogeneity when removing outliers using the leave-one-out method and a Baujat plot (Table 2; Supplementary Figures 2, B; 3, B; and 4, B, available online). Sensitivity analyses for the pooled complete response proportion at 6 months and pregnancy proportion are also reported in Table 2.

Table 2.

Sensitivity analysis for the pooled complete response rate and the pregnancy rate^a

	Oral progestin			Levonorgestrel intrauterine device
	Proportion (95% CI)	I2 (95% CI)	Pb	Proportion (95% CI)	I2 (95% CI)	Pb
Primary outcome, the best complete response proportion within 12 mo
Overall estimated pooled complete response	0.82 (0.69 to 0.91)	58 (27 to 76)	<.001	0.95 (0.81 to 0.99)	79 (58 to 89)	<.001
Leave-one-out	0.84 (0.72 to 0.92)	42 (0 to 68)	<.001	0.96 (0.91 to 0.99)	21 (0 to 64)	.10
Removal of outliers	0.86 (0.75 to 0.92)	25 (0 to 60)	<.001	0.96 (0.91 to 0.99)	21 (0 to 64)	.10
Secondary outcome, complete response proportion at 6 mo
Overall estimated pooled complete response	0.68 (0.38 to 0.88)	86 (72 to 93)	<.001	0.92 (0.39 to 0.99)	77 (38 to 92)	<.001
Leave-one-out	0.76 (0.57 to 0.88)	60 (2 to 84)	.02	0.95 (0.68 to 0.99)	0 (0 to 90)	.08
Removal of outliers	0.76 (0.57 to 0.88)	60 (2 to 84)	.02	NA	NA	NA
Secondary outcome, pregnancy
Overall estimated pooled rate	0.50 (0.35 to 0.65)	48 (0 to 78)	.06	0.35 (0.23 to 0.49)	0 (0 to 79)	.64
Leave-one-out	0.55 (0.39 to 0.70)	26 (0 to 69)	.22	0.34 (0.20 to 0.50)	0 (0 to 85)	.58
Removal of outliers	NA	NA	NA	NA	NA	NA

a

The pregnancy rate was defined as the number of patients who were able to conceive divided by the number of patients who attempted to become pregnant. CI = confidence interval; NA = not applicable.

^bP value was tested by likelihood ratio test.

We performed a sensitivity analysis to estimate the effect of levonorgestrel-releasing IUD with or without GnRH agonist. The pooled best complete response proportion within 12 months following levonorgestrel-releasing IUD treatment with a GnRH agonist was 0.96 (95% CI = 0.78 to 0.99), while the proportion without a GnRH agonist was 0.93 (95% CI = 0.59 to 0.99). After removing an outlier study (27), the proportion without a GnRH agonist was 0.96 (95% CI = 0.79 to 0.99) with reduced heterogeneity (Supplementary Figure 5, available online).

Publication bias and risk of bias

Funnel plots for the studies included in the analysis of the primary outcome are presented in Supplementary Figure 6 (available online). Publication bias was seen among studies of oral progestin (Egger test, P = .03), while no evidence of publication bias was seen among studies of levonorgestrel-releasing IUD (Egger test, P = .36). Among studies with the levonorgestrel-releasing IUD, Peter test also showed the same result (P = .16). The risk of bias assessment is summarized in Supplementary Tables 1 and 2 (available online). The overall score was also described to compare the quality of each study. The score ranged 6-9 of 11. The main factors that lowered scores were item 1 (“Were the 2 groups similar and recruited from the same population?”), item 2 (“Were the exposures measured similarly to assign people to both exposed and unexposed groups?), item 5 (“Were strategies to deal with confounding factors stated?”), and item 10 (“Were strategies to address incomplete follow up utilized?”) (Supplemental Table 1, available online).

Discussion

In our meta-analysis of 21 studies, which included 824 premenopausal patients with endometrial intraepithelial neoplasia, 82% of patients treated with oral progestins achieved a complete response within the first 12 months, while 95% of patients treated with levonorgestrel-releasing IUD achieved a complete response within the same time frame. Sensitivity analyses reinforced these findings, indicating consistent effect sizes with reduced heterogeneity. These data suggest that the use of levonorgestrel-releasing IUD for patients with endometrial intraepithelial neoplasia seeking to preserve fertility may offer a substantial chance of treatment success in the first year following treatment than oral progestins. However, these data need to be interpreted with caution given that this meta-analysis was synthesized with a limited number of patients.

Oral progestins have a long history of use for the treatment of endometrial intraepithelial neoplasia, with substantial evidence supporting their efficacy. A previous meta-analysis on patients with complex endometrial hyperplasia with atypia, a closely related pathological definition of endometrial intraepithelial neoplasia (46), included only 117 patients across studies treated with oral progestins and reported a pooled complete response rate of 74% (95% CI = 65% to 81%) (15). Another similar meta-analysis reported a pooled complete response proportion of 69% (95% CI = 58% to 83%) from a subgroup of 189 patients with atypical hyperplasia treated with oral progestin (47). These earlier meta-analyses were limited by a smaller number of available studies at that time and thus included fewer patients than our study. The lower response rate observed in these earlier meta-analyses may have been influenced by a few studies that showed especially low complete response rates (43,48,49). Both prior meta-analyses included pre- and postmenopausal women and included a wide range of assessment times beyond 1 year following treatment. In contrast, our study included a larger number of studies (n=21) and patients (n=824). Although we focused on premenopausal patients, our finding of an 82% complete response rate with oral progestins at 1 year posttreatment is in line with prior studies.

Literature on the efficacy of levonorgestrel-releasing IUD treatment for patients with endometrial intraepithelial neoplasia is more limited. One meta-analysis reported a pooled complete response rate of 90% (95% CI = 62% to 100%) among patients with endometrial intraepithelial neoplasia (36 patients) from 7 studies (47). Conversely, 1 study showed no statistically significant difference in the complete response proportion between patients treated with oral progestins or levonorgestrel-releasing IUD (12), although this study encompassed a mixed cohort of patients with endometrial intraepithelial neoplasia and endometrial cancer. A meta-analysis by Yuk et al. (50) suggested that treatment with levonorgestrel-releasing IUD had a higher efficacy than treatment with oral progestins, however, they included only 2 RCTs, and the patients in these trials included endometrial intraepithelial neoplasia and hyperplasia without atypia (11,51). A Cochrane review incorporating 11 RCTs evaluating the effectiveness and safety of levonorgestrel-releasing IUD treatment reported improved regression of endometrial hyperplasia within 6 months in pre- and postmenopausal patients (OR = 2.94, 95% CI = 2.10 to 4.13) compared with nonintrauterine progestins (10). However, the patient populations across the RCTs were not consistent and included patients with endometrial hyperplasia with and without atypia (11,52), patients with simple endometrial hyperplasia only, or patients with simple or complex nonatypical hyperplasia (10). Only endometrial hyperplasia with atypia is a lesion that has malignant potential as a precursor of endometrial cancer, which indicates this review cannot be directly extrapolated to our findings.

Given the risk of untreated or unresponsive endometrial intraepithelial neoplasia progressing to endometrial cancer, it is crucial to discuss alternative treatment options, including hysterectomy, if complete response is not achieved after 9-12 months of progestational therapy (7,53). In addition, for patients desiring conception, it is recommended to try to conceive expediently when a complete response is achieved to avoid endometrial intraepithelial neoplasia recurrence (7,53,54). Previous studies have either lacked a clearly defined time period or included a wide assessment time period, when assessing the effect of treatment (12,14,15,50,55-57). As such, high-quality evidence regarding the efficacy of levonorgestrel-releasing IUD treatment for patients with endometrial intraepithelial neoplasia who desire fertility preservation remains limited. Despite this, the 2023 American College of Obstetricians and Gynecologists guideline suggests that intrauterine progestational administration may be associated with a higher rate of disease regression when compared with oral progestins alone in patients with endometrial intraepithelial neoplasia (7). Our data lend further support to this recommendation.

Several adjunctive treatment options to enhance the effectiveness of primary progestational treatment have been reported. The effectiveness of GnRH agonists as adjunctive treatment has been documented in several studies (58). A study looking at the synergistic effects of GnRH agonists with levonorgestrel-releasing IUD or aromatase inhibitors reported a complete response rate of 97% within 10 months among patients with endometrial intraepithelial neoplasia (58). Metformin is considered a promising adjunctive therapy alongside progestational therapy for conservative therapy for endometrial intraepithelial neoplasia because of its antiproliferative effects on endometrial cells (56,59). Some studies found no benefit in the response rate among patients with endometrial intraepithelial neoplasia (56,60), however, 1 RCT concluded that metformin plus megestrol acetate was associated with a higher complete response rate in patients with endometrial intraepithelial neoplasia compared with megestrol acetate alone (44). Transcervical resection is recommended as a combination approach followed by oral progestins or levonorgestrel-releasing IUD in a recent 2023 European guideline (53). However, there is a paucity of data on the effectiveness of transcervical resection for endometrial intraepithelial neoplasia. In our meta-analysis, we reported on the use of several adjunctive therapies including GnRH agonists, metformin, and transcervical resection, which may have affected the results of the 2 primary progestational therapies. However, we conducted a subgroup analysis to examine the impact of GnRH agonists on the effectiveness of levonorgestrel-releasing IUD. Although we were limited by a small number of studies in each GnRH agonist group, we observed no substantial differences in the pooled proportion of the best complete response within 12 months following levonorgestrel-releasing IUD with or without a GnRH agonist. Because of an insufficient number of studies, we were unable to examine GnRH agonists with oral progestins or examine metformin or transcervical resection with levonorgestrel-releasing IUD or oral progestins.

Several limitations of our meta-analysis should be acknowledged. First, our pooled odds ratio comparing the best response rate within 12 months between oral progestin and levonorgestrel-releasing IUD treatments was based on only 3 studies, including a total of 310 patients, which may have contributed to the wide confidence intervals and lack of statistical significance of our effect estimate. Second, there were few studies directly comparing outcomes of oral progestins and levonorgestrel-releasing IUD, thus we report the proportion of complete responses at 12 months for each group. Third, there was statistically significant heterogeneity between studies in terms of sample size and effect size. However, we addressed this by performing several sensitivity analyses to remove outlier studies and observed a similar effect size with lower heterogeneity, showing that our results were robust. Fourth, we identified a limited number of studies including our secondary outcomes, which may have been due to our specific inclusion criteria for patients and our primary outcome. Therefore, we may have missed studies that included our secondary outcomes and introduced selection bias into our secondary outcomes results. The goal of our analysis was to analyze premenopausal patients with endometrial intraepithelial neoplasia, however, although we used an age cutoff of 50 years, we cannot ensure that all included patients were indeed premenopausal. There was likely significant heterogeneity in the study population as well as significant selection bias for inclusion of patients motivated to avoid hysterectomy. As such, our findings may not be generalizable to the larger population of reproductive-aged women with endometrial intraepithelial neoplasia. Lastly, we were unable to examine the influence of adjunct therapies on our outcomes because of a small number of studies including specific adjunct therapies.

Our study also had several notable strengths. First, we conducted a thorough search of the literature utilizing 5 research databases and an RCT database using broad and inclusive search criteria developed by a research librarian. Second, unlike previous meta-analyses, this study was limited to patients aged 50 years or younger with endometrial intraepithelial neoplasia only and did not include patients with endometrial hyperplasia without atypia or early endometrial cancer, who are also indicated for fertility-preserving progestational therapy (12). Many prior meta-analyses or RCTs examined treatment effects in patients with the mixed disease statuses such as endometrial intraepithelial neoplasia and endometrial cancer (12,14,57), endometrial intraepithelial neoplasia and endometrial hyperplasia without atypia (50), and postmenopausal patients (11,50), making it challenging to extrapolate their findings to endometrial intraepithelial neoplasia patients wishing to preserve fertility. Finally, we used a clinically meaningful time period (12 months posttreatment) for assessment of complete response following progestational treatment.

In conclusion, this meta-analysis provides data on the effects of oral progestins and levonorgestrel-releasing IUD treatment within 12 months among reproductive-age patients with endometrial intraepithelial neoplasia. While these data suggest that levonorgestrel-releasing IUD is a reasonable treatment option for endometrial intraepithelial neoplasia, there is limited prospective data directly comparing oral and intrauterine progestational therapy and thus a clear need for further comparative data from a multicenter study. These data may help clinical decision making and counseling when selecting progestational therapies for patients with endometrial intraepithelial neoplasia who desire fertility preservation.

Data availability

The datasets were derived from public sources. The data underlying this study can be shared based on the request to the corresponding author.

Author contributions

Yukio Suzuki, MD (Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Writing—original draft; Writing—review & editing), Jennifer Ferris, PhD (Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Writing—original draft; Writing—review & editing), Ling Chen, MD (Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Writing—original draft; Writing—review & editing), Shayan Dioun, MD (Formal analysis; Investigation; Methodology; Writing—review & editing), John Usseglio, MPH (Conceptualization; Data curation; Investigation; Methodology; Writing—review & editing), Koji Matsuo, MD (Formal analysis; Investigation; Methodology; Writing—review & editing), Xiao Xu, PhD (Investigation; Methodology; Writing—review & editing), Dawn Hershman, MD (Formal analysis; Investigation; Methodology; Writing—review & editing), and Jason D. Wright, MD (Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Project administration; Resources; Writing—original draft; Writing—review & editing).

Funding

No author received a grant for this study.

Conflicts of interest

Dr Wright has received royalties from UpToDate and received research funding from Merck. Dr Suzuki reports receiving payment from The Japan Society for Menopause and Women’s Health (JMWH Bayer Grant 2021), from Honjo International Scholarship Foundation (Honjo-JMSA Scholarship 2022), from Kanzawa Medical Research Foundation (Oversea Research Grant 2022), and from Mitsui & Co (USA), Inc (Mitsui USA-JMSA Scholarship 2023). DH, who is a JNCI Associate Editor and co-author on this paper, was not involved in the editorial review or decision to publish the manuscript. All other authors declare no conflicts of interest.

Acknowledgements

The funder had no role in the design of the study; the collection, analysis, or interpretation of the data; or the writing of the manuscript and decision to submit it for publication.

Author notes

Yukio Suzuki and Jennifer S. Ferris contributed equally to this work.