Risk factors of progression to endometrial cancer in women with endometrial hyperplasia: A retrospective cohort study.

Endometrial hyperplasia (EH) is a pathological condition characterized by proliferation of endometrial glandular and stromal structures. The revised World Health Organization (WHO) classification divides EH into non-atypical EH and atypical EH/endometrioid intraepithelial neoplasia without the previous simple and complex subtypes [ 1 , 2 ]. EH, particularly with atypia, is a precursor to endometrial carcinoma (EC) [ 3 ]. In a study in the United States (US), the peak incidence of EH was 142/100,000 and 213/100,000 woman-years in simple and complex non-atypical EH, respectively (in subjects in their early 50s) and 56/100,000 woman-years in atypical EH (in subjects in their early 60s) [ 4 ]. Endometrial cancer (EC) is the most common cancer of the female reproductive tract [ 5 ]. The incidence of endometrial cancer has increased globally due to the increasing number of elderly people and increasing rates of obesity [ 6 , 7 ]. In a retrospective study in which 170 women with EH were followed for a mean of 13.4 years (from 1 to 26.7 years), progression to EC occurred in 1.6% and 23% of women with non-atypical and atypical EH, respectively [ 2 ]. In a matched case-control study, cumulative risk of progression to EC at years 4, 9, and 19 after EH diagnosis was 1.2%, 1.9%, and 4.6%, respectively, in women with non-atypical EH and 8.2%, 12.4%, and 27.5% in women with atypical EH [ 8 ]. Several studies have reported that 10–59% of women with atypical EH had occult EC detected at hysterectomy [ 9 ]. Advanced age, menopause, obesity, diabetes mellitus, abnormal uterine bleeding, and (complex) atypical EH have been reported as predictive factors of concurrent EC in women with EH [ 9 – 11 ]. However, clinical risk factors related to progression to EC in women with non-atypical and atypical EH have not been reported. Some studies have reported that serum DNA integrity index and molecular markers and immune cells related to the immune escape mechanisms may play roles in EC [ 12 , 13 ]. Evaluating these roles in EH might be of benefit to predicting progression to EC in women with EH. Management for EH has aimed to control symptoms such as heavy bleeding, detect concurrent EC, and prevent subsequent development of EC [ 14 ]. However, no studies have evaluated risk factors that predict progression to EC in women with EH. Therefore, this study aimed to investigate factors that influence EH progression to EC.

Data from 62,333 women were entered into the database, and the data from 13,744 women met the selection criteria ( Fig 1 ). The data from 11,525 women with non-atypical EH and 2,219 women with atypical EH were extracted for analysis. In total, 48.2% of women with EH were diagnosed between 40–49 years of age. Younger and older ages had lower incidences of EH. Age at diagnosis of EH was not different between women with non-atypical and atypical EH. Type 2 diabetes and endometriosis increased in women with atypical EH compared to those with non-atypical EH. In addition, 30.9% of women with EH did not receive follow-up biopsy after diagnosis of EH, and the frequency of this result was similar in the 2 EH types. Of note, 34.6% of those subjects received only 1 follow-up biopsy, and the frequency was higher in women with non-atypical EH. 3 and ≥4 follow-up biopsies were performed more frequently in women with atypical EH. Abnormal uterine bleeding, tamoxifen use, and use of hormone therapy (LNG-IUS and progestational agents) were not different between women with non-atypical and atypical EH ( Table 1 ). Data for oral contraceptive use were not available. EH, endometrial hyperplasia a Levonorgestrel-releasing intrauterine system Incidence of EC showed a higher tendency in women with atypical EH than in those with non-atypical EH. The C54.1 diagnostic code of endometrial EC was identified in 88.7% of EC cases. Incidence of EC diagnosed with C54.1 was not different between EH types ( Table 2 ). In 1< and ≤ 3 years after diagnosis of EH, EC occurred more frequently in women with non-atypical EH compared to those with atypical EH. EC occurred more frequently in women with atypical EH 3 years after diagnosis of EH than for those with non-atypical EH. However, in cases diagnosed at 5< and ≤ 6 years, EC occurred similarly in both non-atypical and atypical EH. Moreover, contrary to atypical EH, in women with non-atypical EH, EC occurred most frequently at 1< and ≤ 3 years after diagnosis of EH and decreased thereafter ( Table 2 ). EH, endometrial hyperplasia; EC, endometrial cancer Both women with non-atypical and atypical EH showed significant decrease in risk of EC at 40–49 years old compared to women ≤39 years. However, risk of EC was not significantly different between women ≤39 and those ≥50 years old; the incidence of EC increased in women with type 2 diabetes. However, type 2 diabetes was not a risk factor for EC; risk of EC tended to decrease in women with endometriosis when adjusted for other confounding factors ( Table 3 ). EH, endometrial hyperplasia; EC, endometrial cancer; HR, hazard ratio; CI, confidence interval a Levonorgestrel-releasing intrauterine system. b The data were adjusted for all risk factors (age at diagnosis of EH, type 2 diabetes, endometriosis, abnormal uterine bleeding, use of tamoxifen, LNG-IUS, progestational agents, and number of follow-up biopsies after diagnosis of EH). Women with non-atypical EH showed increase in EC in women with abnormal uterine bleeding and LNG-IUS. However, there were no risk factors for EC when adjusted for other confounding factors, and the incidence of EC increased in women using tamoxifen. However, tamoxifen use was not a risk factor for EC; progestational agents were associated with significant decrease of EC risk when adjusted for other confounding factors, and the risk of EC significantly increased according to number of follow-up biopsies after diagnosis of EH ( Table 3 ). The following findings were identified for women with atypical EH: abnormal uterine bleeding, tamoxifen use, and use of hormone therapy (LNG-IUS and progestational agents) were not associated with EC and were not risk factors for EC, and the risk of EC significantly increased when ≥2 follow-up biopsies were performed after diagnosis of EH ( Table 3 ). The cumulative incidence of EC was associated with a higher tendency in women with atypical EH than in those with non-atypical EH ( P = 0.082). The incidence density of EC was not significantly different between women with non-atypical and atypical EH ( P = 0.913). Cumulative incidence and incidence density of EC according to age at diagnosis of EH were associated with the following findings: in women with non-atypical EH, the lowest incidence occurred in women 40–49 years of age and the highest in women ≥70 years old; in women with atypical EH, the lowest incidence was for women 40–49 years of age and the highest was in women 30–39 or ≥70 years of age ( Table 4 ). EH, endometrial hyperplasia; EC, endometrial cancer Time to progression to EC in women with non-atypical EH decreased in women ≥50 years old compared to other ages (≤39 and 40–49 years old). Time to progression to EC in women with atypical EH was not different among ages ( Table 4 ).

Based on claim data analysis, we demonstrated that, regardless of EH type, women ≤39 and ≥50 years of age with EH had high risk for progression to EC. Moreover, this study indicates the importance of repeated follow-up biopsy after diagnosis of EH regardless of EH type. Finally, different influences of progestational agents for progression to EC depending on EH type may support current trends for non-surgical management in women with non-atypical EH and surgical management in women with atypical EH.

South Korea has a universal health coverage system, the National Health Insurance, that covers approximately 98% of the overall Korean population. The claims data of the Health Insurance Review and Assessment Service (HIRA) represent 46 million patients per year [ 15 ]. This claims data study used the HIRA databases and data with EH diagnostic codes generated between January 1, 2007 and February 28, 2018. Inclusion criteria were women who had diagnostic codes for EH with procedure codes for endometrial biopsy within 60 days before or after an initial diagnostic code and women receiving 1 or more management approaches for EH at least 90 days after diagnosis of EH. Women who were diagnosed with EC or who underwent hysterectomy within 1 year after diagnosis of EH were excluded. Because the HIRA dataset uses anonymous identification codes to protect patients' personal information, approval of this study was waived by the Institutional Review Board of Inha University Hospital (No. 2019-11-015) on November 25, 2019. The diagnostic codes in the 10th revision of the International Statistical Classification of Diseases and Related Health Problems (ICD-10) were used to obtain data for women who had been diagnosed with EH (N85.0: endometrial glandular hyperplasia and N85.1: endometrial adenomatous hyperplasia [atypia]). Simple or complex EH was not distinguished. The procedure codes were derived from health insurance medical care expense claim forms. The procedure codes for endometrial biopsy were dilatation and curettage (R4521), endometrial biopsy (C8571, C8572), aspiration biopsy (C8573), simple curettage (C8574), and hysteroscopic curettage (C8575). The procedure codes for hysterectomy were simple abdominal hysterectomy (R4143, R4147), complex abdominal hysterectomy (R4144, R4148), laparoscopic hysterectomy (simple [R0141] and complex [R0142]), subtotal hysterectomy (R4130), vaginal hysterectomy (simple [R4149] and complex [R4140]), and radical hysterectomy (R4154, R4155). EC was assigned in women who had related diagnostic codes (ICD-10: C54, C54.0, C54.1, C54.2, C54.3, C54.8, C54.9, C55) with procedure codes for endometrial biopsy or hysterectomy within 60 days before or after an initial diagnostic code or women who had diagnostic codes for EC more than 2 times within 1 year. In Korea, every person with a cancer diagnosis is registered with a unique code (called a C code) in the National Cancer Registry. This C code is used in all subsequent medical records and claims created for that patient. Therefore, cancer diagnosis based on claims is considered reliable [ 16 ]. Type 2 diabetes was defined as the presence of identical E11-E14 codes (ICD-10) at least twice for 1 patient or a diabetes drug code (including biguanides, sulfonylurea, meglitinides, thiazolidinediones, dipeptidyl peptidase-4 inhibitors, α-glucosidase inhibitors, sodium-glucose co-transporter 2 inhibitors, insulin, or glucagon-like peptide 1 agonists) plus an E11-E14 code [ 16 ]. Endometriosis was defined as presence of a diagnostic N80X code (ICD-10) with associated procedures that included fulguration (R4165), ovarian cystectomy (R4421, R4430), adnexectomy (R4331, R4332), and hysterectomy (R4147, R4148, R0141, R0142, R4130, R4149, R4140) within 60 days before or after an initial diagnostic code. ICD-10 codes for abnormal uterine bleeding were N93, N93.8, and N93.9. Progestational agents comprised medroxyprogesterone acetate and megestrol acetate. All data collection was performed in parallel for both non-atypical and atypical EH. EH type; age at diagnosis of EH; presence of type 2 diabetes, endometriosis, or abnormal uterine bleeding; a diagnostic code for EC; tamoxifen use; types and use of hormone therapy (levonorgestrel-releasing intrauterine system [LNG-IUS] or progestational agents); number of follow-up biopsies after diagnosis of EH; and time from diagnosis of EH to diagnosis of EC were extracted. SAS ® Enterprise Guide ® version 6.1 (SAS Institute, Inc., Cary, NC, USA) was used for data mining and analysis. Categorical variables were reported as number and percentage, and continuous variables were reported as mean ± standard deviation (SD). The categorical variables were analyzed using the Chi-square test or Fisher’s exact test, whereas the continuous variables were analyzed using the two-tailed independent t test or one-way analysis of variance (ANOVA) followed by Bonferroni's correction. In addition, the associations of variables with EC in each EH type were analyzed using the Cox Proportional Hazard Regression model with or without adjusting for confounding factors. All variables were used as confounding factors. P values <0.05 were considered statistically significant.