Data source

After Institutional Review Board approval was obtained at the University of Southern California, a retrospective observational cohort study was performed including all consecutive cases of CAH. First, medical records of women who had a histopathologic diagnosis of any type of endometrial hyperplasia between 2003 and 2018 at Los Angeles County USC Medical Center were reviewed. These records were obtained from pathology database records including all endometrial sampling or surgical specimens with pathology reports that contained the words “endometrial” and “hyperplasia.” During the study period, the WHO 1994 classification system was used on pathology reports and the newer endometrial intraepithelial neoplasia schema was not implemented in our institution.

Pathology records were then linked with clinical medical records. Patients diagnosed early in the study period were also included in our prior study. Data entry was performed by 5 investigators. Upon completion of the data collection, one fifth of the data was randomly reviewed for accuracy, consistency, and integrity of data entry by 3 investigators (R.S.M, M.A.C, and K.M.). Data review was censored at that point if there was high accuracy in the data entry.

Eligibility

Women with CAH who received medical treatment with progestin were eligible for the study criteria. Women with nonatypical or simple hyperplasia as well as those with an endometrial cancer diagnosis prior to the start of progestin therapy were excluded. Of women with CAH, only those who had received at least 1 month of systemic or local progestin therapy following the diagnosis of CAH and who did not initiate progestin therapy before CAH diagnosis were included. Patients on multiple progestin agents were also excluded. Women were excluded if they did not have a follow-up biopsy performed at least 1 month following treatment initiation.

Clinical information

Patient demographic variables included age at CAH diagnosis, race/ethnicity, gravidity, parity, body mass index (BMI, kg/m ² ), medical comorbidities (diabetes mellitus, hypertension, hyperlipidemia, polycystic ovary syndrome, reported infertility, and personal cancer history), and medication use (metformin, aspirin, statins, and beta blockers). Endometrial thickness on ultrasonographic imaging at the time of CAH diagnosis was also recorded.

Histopathologic diagnosis and date of procedure were retrieved for all endometrial tissue sent to pathology for analysis, including samples from endometrial biopsies, suction or sharp dilation and curettage, and hysterectomy specimens. Progestin route of administration was recorded, including oral medroxyprogesterone acetate, megestrol acetate, or norethindrone, intramuscular depo-medroxyprogesterone acetate, or the LNG-IUD. Date of treatment initiation and last follow-up date were recorded. Chronology of CAH diagnosis, progestin treatment initiation, and outcome on follow-up biopsies were then ascertained from this information. Among those who developed subsequent endometrial cancer, tumor characteristics, treatment type, and survival outcomes were abstracted.

Treatment follow-up

At our institution, complex hyperplasia on biopsy is managed with suction dilation and curettage, followed by initiation of medical management or hysterectomy. This practice is to ensure that there is no underlying malignancy given the high rate of occult malignancy in CAH, as previously described. Resampling of endometrial tissue is typically performed every 3–6 months in those who elect for medical management to assess treatment response. In those who achieve complete response, follow-up is generally continued in the same fashion but may be individualized depending on specific patient risk factors or clinical considerations.

Definition of grouping allocation

Body habitus was grouped by the Centers for Disease Control and Prevention criteria as follows: nonobese (BMI <30 kg/m ² ), class I obesity (BMI 30–34.9 kg/m ² ), class II obesity (BMI 35–39.9 kg/m ² ), and class III obesity (BMI ≥40 kg/m ² ). Progestin therapy was grouped as local therapy vs systemic therapy: local therapy referred to the LNG-IUD and all others were classified as systemic therapy in this study. Histopathology samples were classified based on the 1994 WHO criteria as follows: (1) simple hyperplasia without atypia, (2) simple hyperplasia with atypia, (3) complex hyperplasia without atypia, (4) complex hyperplasia with atypia, (5) cancer, and (6) no hyperplasia identified. ^,

Treatment response was categorized as follows: (1) complete response, defined as no residual hyperplasia identified on subsequent biopsies; (2) partial response, defined as regression of CAH to simple or nonatypical hyperplasia; (3) persistent CAH; and (4) progression to cancer. Treatment outcomes were then classified for statistical analyses as complete response, overall response (complete or partial response), persistent CAH, or progression to cancer. Time to complete or partial response was defined as the time interval between progestin therapy initiation and the date of the first follow-up biopsy showing treatment response. Time to cancer progression was defined as the time interval between progestin therapy initiation and date of endometrial cancer diagnosis. The last follow-up date was defined as the last biopsy date, and the cases were censored at the last follow-up visit if an outcome event was absent or at the time point when the patient received a second hormonal agent or underwent hysterectomy.

Statistical analysis

The goal of the primary analysis was to compare treatment response between patients who received the LNG-IUD and patients who received systemic therapy for CAH. In the secondary analyses, we examined patient characteristics and treatment patterns among women with CAH who received medical management.

Differences in patient demographics between the 2 groups were assessed with the Student t test, Fisher exact test, or χ ² as appropriate. Because treatment response to progestin therapy depends on duration of follow-up, a time-dependent analysis was performed. Cumulative response curves (or time to progression to cancer) were plotted with Kaplan-Meier methods, and differences in curves were assessed with the log-rank test. Cox proportional hazard regression models were then used to examine associations between progestin therapy route and treatment response (complete response, overall response, or cancer progression). The magnitude of statistical significance was expressed with adjusted hazard ratios (HR) and 95% confidence intervals (CI).

Propensity score (PS)-based inverse probability of treatment weighting (IPTW) was used to balance the background differences between the 2 groups. The IPTW model creates a weighted cohort that differs based on treatment type (LNG-IUD vs systemic therapy) but is similar with respect to other baseline demographics. First, the PS was established by fitting a binary logistic regression model. Owing to the small sample size, effect size was used for the selection of covariates, and all the covariates with standardized difference [SD] ≥0.20 were entered in the model. The IPTW approach assigned women who received the LNG-IUD a weight of 1/PS and those who received systemic therapy a weight of 1/(1 − PS). Stabilized weights were used in the analysis, and the threshold technique was used to trim weights at the first and 99th percentiles. In the weighted model, distributions of demographic variables were assessed, and an SD ≤0.10 was considered a good balance between the 2 groups.

Various sensitivity analyses were performed to examine the robustness of our analysis. First, an interaction-term analysis was performed to examine the impact of body habitus and treatment response. This was based on the hypothesis that women with a larger BMI are less likely to respond to progestin therapy and more likely to progress to cancer. Complete response rates were calculated for each BMI category (nonobesity, class I–II obesity, and class III obesity). Second, doubly robust estimator was used to correct for possible confounding variables for outcome. Parsimonious models were fitted by adjusting for factors (age, BMI, and diabetes status) that are known to affect CAH treatment response and time factor. ^{, ,}

Last, a cohort-level analysis was performed to test the hypothesis that increasing LNG-IUD use was associated with increased complete response rates in the study cohort. The trend in LNG-IUD use over time was first plotted, and linear segmented regression with log-transformation was used to assess for a time point at which trends in LNG-IUD use significantly changed. The study period was then divided at this time point. A quasi-experimental approach with an interrupted time-series analysis was employed to assess the complete response rates in the separate segments before and after this time point, and the results corresponding to the LNG-IUD increasing period was interpreted as treatment effect.

All statistical analyses were based on 2-sided hypotheses, and a P value of less than .05 was considered statistically significant. Statistical Package for Social Sciences (version 24.0; SPSS, Armonk, New York) and “IPWsurvival” package in R version 3.5.0 (R Foundation for Statistical Computing, Vienna, Austria) were used for all analyses. The STROBE guidelines were consulted to outline this observational cohort study.